OUT-OF-FIELD TEACHING: A CROSS-NATIONAL STUDY ON TEACHER LABOR

MARKET AND TEACHER QUALITY

by

Yisu Zhou

A DISSERTATION

Submitted
to Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Educational Policy

2012
 ABSTRACT

OUT-OF-FIELD TEACHING: A CROSS-NATIONAL STUDY ON TEACHER LABOR

MARKET AND TEACHER QUALITY

by

Yisu Zhou

In the past two decades, the issue of out-of-field teaching (OFT) has concerned pol-
icy makers and researchers alike who see raising teachers’ subject matter knowledge as the
main policy lever to improve teacher quality. The study of OFT has emerged as one of the
important subfields of teacher quality and teacher labour market research. Researchers in
the United States have found widespread and high reliance on out-of-field teachers at the
secondary level. They argue that out-of-field teachers are so great in number that the neg-
ative influence on student learning is spreading. Researchers have argued that OFT could
undermine efforts to improve teacher quality through the professionalization of the teach-
ing occupation. Outside the United States, studies have found that out-of-field assignments
exist in several countries. However, OFT as a phenomenon remains understudied in the in-
ternational comparative literature. We are unaware of the exact nature and spread of OFT
internationally. This study examines OFT from an international comparative perspective
using the newly published data from OECD. I focus on math and science teachers who teach
in public schools in 21 countries. Three questions are examined in this study. First, is there
cross-national variations in out-of-field teaching? Second, what is the distribution of out-
of-field teachers across schools? Third, are there any differences in teaching practices and
received professional support between out-of-field and in-field teachers? I find out-of-field
teachers share certain similar attributes: young and inexperienced teachers with substan-
tial educational attainment who work on short contract and part-time bases. Out-of-field
teachers are disproportionally concentrated in rural, small, and low-SES schools. While out-
of-field teachers do not differ from in-field teachers in several measures of teaching practices
and time allocation, they were not given enough on-the-job development opportunities to
improve their teaching skills.
Copyright by
YISU ZHOU
2012
 ACKNOWLEDGEMENTS

This dissertation is the intellectual product from the six years I spent at Michigan State
University, together with the previous experience I gained as a rural school teacher in China.
Throughout the years, this project has benefited a great deal from many people I worked
with and encountered at different stages of my life. I cannot express enough of my gratitude
here but I would like to take the opportunity to acknowledge their support. Prof. Amita
Chudgar has spent a great mount of time to help me shape the course of this dissertation. I
was fortunate to have Prof. Chudgar as my graduate advisor. She has been a great mentor
all along the way and a great motivator to make many of the things I presented in this
dissertation the way they are right now. Not only have I gained many professional insights
from her about my graduate career; she has also shown me an admirable model of diligent
work and top-notch research skills. I particularly appreciate her patience and kindness to
many of the questions I raised during the years. For an international student studying in a
unfamiliar place, nothing warms my heart more than a caring adviser. Choosing her as my
adviser was the best choice during my PhD study.
I also express my thanks to my family. My parents, Zhou Jianming and Gu Guangqing
were my first teachers in academics even before I started graduate school. Their uncondi-
tional love made me who I am today. They have shown me that scholarly work is rooted
in the connection with the people and society surrounding you. I am continuously inspired
by their deeds. My lovely wife, Wan Biru, has been my closest friend since high school and
has continued to support my venture in academia during the past six years. She is always a
place of warm comfort and has stood by me throughout every difficulty. I have found great
strength from her. My friends from childhood, Yan Liang, Qian Jun, Lu Qing-hua, Yan Min,
Lin Xiaoxi, Hu Jiachen, and Ren Min have brought me much joy over the years. Though
working in various fields and living in different places, they always bring new perspectives

v
to my life. I am so glad I will be able to join them once again soon.
I have spent many good times in East Lansing. During my study at MSU, I encountered
many wonderful people. I would like to thank Ren Zhen, Wen Hao, Zheng Zhen, Fu Jing, Li
Ning, Wang Chen, Tao Zhengsheng, Xu Kaijie, Lin Yiyan, Chen Zhong, and Huang Yi-jen
for making my life colorful in this beautiful place. I wish them all good luck on their paths.
The people I interacted with on daily basis, my colleagues at the College of Education at
MSU, are of great importance to me. I benefited from working with them. Wang Qiu, Sun
Min, Hsieh Chueh-an, Lo Yun-Jia, and Seung-Hwan Ham, who are senior in the program to
me have proved to be great source of learning. Many of them helped me during my graduate
study and I look forward to collaborate with them in the future. My fellow students from
the education policy program, measurement and quantitative methods program, and teacher
education program have also been great company in graduate school. I thank Justina Judy,
Juanita Bautista Guerra, Liu Na, Li Wei, Guan Saw, Madhur Chandra, Todd Drummond,
Dave Dai, and Sedat Gumus. I wish them all the best in their graduate study and subsequent
life.
The faculty members that also participated in this project, Prof. Peter Youngs, Prof.
Richard Houang, and Prof. David Arsen, have provided much help in the past years. To-
gether they have shown me great intellectual insights from different areas. Their critical
feedback has always been source of learning to me. I thank them for the rigor they brought
me and my study. I would like to express additional gratitude to Peter for offering me great
advice in navigating professional career.
I specially thank two faculty from MSU. Prof. Barbara Schneider has provided me with
a great job for the past five years. I have worked under her as a graduate assistant and she is
a wonderful supervisor. Together we have done several projects and I am looking forward to
continue my work with her in the future. Prof. Schneider has nurtured me with her excellent
scholarship. She has also shown me how to lead an active and fruitful academic life. Prof.
Michael Sedlak has offered the best graduate courses I have ever taken. I have been lucky

vi
to take two courses with him over the years. The historical understanding he helped me to
develop will benefit me for a very long time. I enjoy every conversation I have with him.
I also appreciate the resources he invested in me as the associate dean of the College of
Education.
Lastly, I thank Christina Mazuca for her excellent supportive work. I cherish her friend-
ship and wish her a wonderful life. Leigh Ann Halas has provided some excellent editing
work. I thank her for making this dissertation the way it is. Other people I worked with,
Prof. Spyros Konstantopoulos from MSU, Prof. Thomas Luschei from Claremont Graduate
University, and Ms. Yang Hongyu from the World Bank, have all shown me different aspects
of great professional work. I also wish them the best. I could not list all the people who
have granted their support to me, but give my heartfelt thanks to every one of them.

vii
 TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii

CHAPTER 1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . 1

CHAPTER 2 CONCEPTUAL FRAMEWORK AND REVIEW OF REL-

EVANT LITERATURE . . . . . . . . . . . . . . . . . . . . . . 7
2.1 Theoretical framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.1 Conceptualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.2 The origins of out-of-field teaching . . . . . . . . . . . . . . . . . . . 10
2.2 Review of relevant literature . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.1 Measuring out-of-field teaching . . . . . . . . . . . . . . . . . . . . . 14
2.2.1.1 Out-of-field in main assignment . . . . . . . . . . . . . . . . 15
2.2.1.2 Out-of-field by subject area taught . . . . . . . . . . . . . . 17
2.2.1.3 Out-of-field by class taught . . . . . . . . . . . . . . . . . . 17
2.2.1.4 Student taught in-field or out-of-field . . . . . . . . . . . . . 18
2.2.2 Spread of out-of-field teaching . . . . . . . . . . . . . . . . . . . . . . 18
2.2.3 Effect of out-of-field teaching . . . . . . . . . . . . . . . . . . . . . . 27
2.2.4 Limitation of the literature . . . . . . . . . . . . . . . . . . . . . . . . 29

CHAPTER 3 RESEARCH QUESTIONS AND DATA DESCRIPTION . . 32

3.1 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2.1 Measuring out-of-field teaching . . . . . . . . . . . . . . . . . . . . . 41

CHAPTER 4 ATTRIBUTES OF OUT-OF-FIELD TEACHERS . . . . . . 45

4.1 The distribution of out-of-field teachers across countries . . . . . . . . . . . 46
4.2 Demographic background of out-of-field teachers . . . . . . . . . . . . . . . . 52
4.2.1 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.2.2 Age . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2.3 Employment status . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2.4 Contract length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2.5 Educational attainment . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.2.6 Teaching experience . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.3 Profiles of out-of-field teachers in selected countries . . . . . . . . . . . . . . 74
4.3.1 Country profile: Australia . . . . . . . . . . . . . . . . . . . . . . . . 75
4.3.2 Country profile: Mexico . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.3.3 Country profile: Belgium (Flemish) . . . . . . . . . . . . . . . . . . . 78
4.3.4 Country profile: Norway . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.3.5 Country profile: Turkey . . . . . . . . . . . . . . . . . . . . . . . . . 80

viii
 4.4 Characteristics of schools where out-of-field teachers work . . . . . . . . . . . 82
4.4.1 School location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.4.2 School size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.4.3 School socioeconomic (SES) status . . . . . . . . . . . . . . . . . . . 89
4.4.4 Principal leadership . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.4.5 Teacher shortage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

CHAPTER 5 CROSS-NATIONAL VARIATION IN OUT-OF-FIELD TEACH-

ING ASSIGNMENTS AND THE DIFFERENCES BE-
TWEEN OUT-OF-FIELD AND IN-FIELD TEACHERS . . 100
5.1 Systemic factors explaining cross-national variation in out-of-field teaching . 100
5.1.1 Empirical model specification . . . . . . . . . . . . . . . . . . . . . . 105
5.1.2 Data description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.1.3 Model output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.2 Comparing teaching behavior and professional status between out-of-field
and in-field teachers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.2.1 Empirical model specification . . . . . . . . . . . . . . . . . . . . . . 129
5.2.2 Data description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.2.3 Model output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.2.3.1 Time on task . . . . . . . . . . . . . . . . . . . . . . . . . . 136
5.2.3.2 Disciplinary climate . . . . . . . . . . . . . . . . . . . . . . 139
5.2.3.3 Teaching practice . . . . . . . . . . . . . . . . . . . . . . . . 142
5.2.3.4 Professionl status . . . . . . . . . . . . . . . . . . . . . . . . 145
5.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

CHAPTER 6 DISCUSSION AND CONCLUSION . . . . . . . . . . . . . . 151

6.1 Summary of findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
6.2 Implication of policy and administration . . . . . . . . . . . . . . . . . . . . 153
6.3 Limitation of this study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
6.3.1 Limitation in conceptualization of out-of-field teaching . . . . . . . . 158
6.3.2 Limitation of data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
6.4 Recommendation for future TALIS studies . . . . . . . . . . . . . . . . . . . 162

APPENDIX A TECHNICAL NOTES . . . . . . . . . . . . . . . . . . . . . . 167

A.1 Statistical computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
A.2 Typesetting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

APPENDIX B ADDITIONAL TABLES . . . . . . . . . . . . . . . . . . . . . 168

REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171

ix
 LIST OF TABLES

2.1 Percentage of U.S. public school teachers who taught high school without an
undergraduate or graduate major in their main assignments in selected subjects,
1987-88 to 2007-08 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.2 Percentage of 8th -grade students in TIMSS whose teachers’ area of study is not
in math or science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.1 Participatant Countries in TALIS 2008 . . . . . . . . . . . . . . . . . . . . . . . 39

3.2 Description of samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

3.3 Description of variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

4.1 Age distribution of male and female OFT . . . . . . . . . . . . . . . . . . . . . 55

4.2 Age distribution of male and female IFT . . . . . . . . . . . . . . . . . . . . . . 56

4.3 Distribution of educational attainment of out-of-field teachers under age 30 . . . 60

4.4 Distribution of teacher experience by two age groups among out-of-field teachers 62

4.5 Contract type and part-time status of out-of-field teachers over age 50 . . . . . . 63

4.6 Teaching experience of advanced degree holders in selected countries . . . . . . . 71

4.7 Age distribution of Turkish teachers by gender and out-of-field status . . . . . . 81

4.8 Description of school-level characteristics . . . . . . . . . . . . . . . . . . . . . . 83

4.9 Variance partitioning of school-level OFT in math & science . . . . . . . . . . . 87

4.10 Number of schools by parental education level . . . . . . . . . . . . . . . . . . . 91

4.11 Description of administrative leadership variable . . . . . . . . . . . . . . . . . . 95

5.1 Description of independent variables . . . . . . . . . . . . . . . . . . . . . . . . 111

5.2 Description of missing variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

5.3 Estimation results from school-level pooled analysis (1) . . . . . . . . . . . . . . 114

5.4 Estimation results from school-level pooled analysis (2) . . . . . . . . . . . . . . 116

5.5 Estimation results from school-level pooled analysis with interaction terms . . . 118

x
5.6 Estimation results from alternative specification (1) . . . . . . . . . . . . . . . . 123

5.7 Estimation results from alternative specification (2) . . . . . . . . . . . . . . . . 125

5.8 Description of variables used as outcomes . . . . . . . . . . . . . . . . . . . . . . 132

5.9 Description of index variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

5.10 Description of missing variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

5.11 Comparing time spent on various tasks between out-of-field and in-field teach-
ers, school fixed-effects approach . . . . . . . . . . . . . . . . . . . . . . . . . . 137

5.12 Comparing time spent on various tasks between out-of-field and in-field teach-
ers, HLM approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

5.13 Comparing classroom management between out-of-field and in-field teachers,

school fixed-effects approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

5.14 Comparing classroom management between out-of-field and in-field teachers,

HLM approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

5.15 Comparing teaching practices between out-of-field and in-field teachers, school
fixed-effects approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

5.16 Comparing teaching practices between out-of-field and in-field teachers, HLM
approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

5.17 Comparing professional status between out-of-field and in-field teachers, school
fixed-effects approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

5.18 Comparing professional status between out-of-field and in-field teachers, HLM
approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

B.1 Number of schools by the size of teaching force in math and science in the
TALIS sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

xi
 LIST OF FIGURES

2.1 Factors and policies that relate to teacher shortage . . . . . . . . . . . . . . . . 13

4.1 Percentage of out-of-field teachers in math and science, unweighted. For inter-
pretation of the references to color in this and all other figures, the reader is
referred to the electronic version of this dissertation . . . . . . . . . . . . . . . . 46

4.2 Percentage of out-of-field teachers in math and science, weighted by part-time

status . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

4.3 Comparison of out-of-field estimates from TALIS 2008 and TIMSS 2007 data
(8th grade) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.4 Percentage of out-of-field teachers by subject fields . . . . . . . . . . . . . . . . 52

4.5 Proportion of male teachers by out-of-field status. . . . . . . . . . . . . . . . . . 53

4.6 Percentage of teachers under age 30 by out-of-field status . . . . . . . . . . . . . 58

4.7 Percentage of teachers over age 50 by out-of-field status . . . . . . . . . . . . . . 59

4.8 Distribution of teachers by part-time status . . . . . . . . . . . . . . . . . . . . 65

4.9 Percentage of teachers who work part-time by out-of-field status . . . . . . . . . 66

4.10 Percentage of teachers with less-than-one-year contract by out-of-field status . . 68

4.11 Percentage of teachers with permanent contract by out-of-field status . . . . . . 69

4.12 Percentage of teachers without a bachelor’s degree by out-of-field status . . . . . 71

4.13 Percentage of teachers with advanced degrees by out-of-field status . . . . . . . 72

4.14 Percentage of teachers with less than 5 years working experience by out-of-field
status . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

4.15 Percentage of teachers with longer than 20 years working experience by out-of-
field status . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

4.16 Characteristics of teachers by out-of-field status in Australia . . . . . . . . . . . 76

4.17 Characteristics of teachers by out-of-field status in Mexico . . . . . . . . . . . . 77

4.18 Characteristics of teachers by out-of-field status in Belgium (Flemish) . . . . . . 79

xii
4.19 Characteristics of teachers by out-of-field status in Norway . . . . . . . . . . . . 80

4.20 Characteristics of teachers by out-of-field status in Turkey . . . . . . . . . . . . 82

4.21 Percentage of math and science teachers who are out-of-field by school location . 85

4.22 Percentage of teachers who are out-of-field by school size . . . . . . . . . . . . . 88

4.23 Proportion of teachers who work in small schools by out-of-field status . . . . . 89

4.24 Percentage of teachers who are out-of-field by school SES . . . . . . . . . . . . . 92

4.25 Percentage of teachers who are out-of-field by classroom SES . . . . . . . . . . . 93

4.26 Percentage of teachers who are out-of-field by principal’s administrative leadership 96

4.27 Percentage of teachers who are out-of-field by school personnel shortages level . 98

5.1 The relationship between country-level OFT with the principal’s administrative
leadership . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

5.2 The relationship between country-level OFT with the autonomy of hiring teachers 103

5.3 The relationship between country-level OFT with the shortages of teachers . . . 104

xiii
 Chapter 1
INTRODUCTION

When I started teaching in Yangguo Township Middle School, a rural village school
located in northwestern China, teaching was very different than I expected. My college
major was Statistics. After graduation, I decided to participate in a public campaign to call
college graduates to serve the western provinces (or “Go west!” as they called it), in which
the harsh natural environment and the stagnated economy severely limited development in
education. I joined a non-governmental organization (NGO) which ran a network of about
one hundred schools. Schools submitted proposals to the network seeking different kinds of
help (libraries, computers, personnel, etc). I went to Yangguo because they specifically said
they needed teachers. I expected to be assigned to teach a subject such as mathematics or
physics. Instead, when I reported to school in the summer of 2005, the principal assigned
me to an English class.
I was puzzled at this decision and asked him why. “Because you are among the very few
who have attended college and also are willing to teach in a rural school,” he said. “We have
enough math and physics teachers, but never have too many good English teachers here.
Our teachers do not speak as fluently as you do. You have learned college-level English,
something most of our teachers could not achieve. I am sure you will be an excellent English
teacher. Don’t worry about your skills.” I was still not convinced, but I had no choice other
than to spend my time maintaining some computers donated to the schools via the NGO. I
really wished to be involved in academic teaching. That was how I started my experience as
a volunteer teacher from Shanghai in a rural classroom.
I had not been formally trained to be a teacher though I attended a normal university1 ;
1 Historically, graduates from a normal university should all enter teaching. But a reform
initiative in the 1990s gave more flexibility to universities. Many normal universities in China
today have transformed into full-fledged research universities, but at the same time have kept
their teacher-training programs. Students attending these institutions are no longer bound

1
neither did the NGO offered me any pedagogical training. I soon discovered that I had
some advantages over my colleagues who were trained in the local teacher’s college or normal
schools. I had good training in the English language up to the college level, which means I had
mastery of more vocabulary and more complex grammar than my colleagues. Perhaps equally
important, I had ample testing experience; I had taken numerous English language tests from
third grade through college, and I had just taken the Graduate Record Examination before
I became a teacher. This skill of understanding the test helped me to get good scores on
tests, which turned out to be useful to my students (and school) later. My colleagues, at
the same time, mostly completed the local normal school’s training at the secondary level,
and some went to provincial teacher training institutions at the tertiary level. They hardly
had any other exposure to English or western culture (movies, media, or direct contact with
foreigners) outside of their training. Yet I also found myself in a somewhat awkward position
when the semester started. When I opened my first class, I had no idea how to manage a
class size of 73; neither did I know how to develop a lesson plan for the content I was going
to cover. I asked for permission to observe my colleagues’ classes, and I was amazed by
how good they were at classroom management and understanding the curriculum. Though I
picked up these essentials pretty fast, I still struggled from time to time with how to convey
what I knew to my sixth graders, who knew nothing about the English language before our
class.
This one-year experience was precious to me, both on a personal level and on professional
grounds. I enjoyed being with my students.2 Most of all, I realized the hard decisions with
which rural schools educators are faced. There is little choice when it comes to staffing
teaching positions. I also started to ponder what really constitutes a good teacher. Though
many of my colleagues in the school did not receive professional training at college level,
some of them were terrific teachers in the classrooms.
to be teachers.
2 A photographic record of this one-year experience can be found here:
http://www.flickr.com/photos/makzhou/sets/918243/

2
 Later, when I was halfway through my doctoral program, I discovered a phenomenon in
American classrooms similar to the situation I encountered during that one-year experience
in China. “Out-of-field teaching” typically refers to teachers who are teaching subjects out of
their field of training. Out-of-field teaching is long-rooted in American schools (Hechinger,
1985, October 8). This phenomenon has been extensively studied since the 1990s, with the
publication of a series of reports from the National Center for Educational Statistics (Bobbitt
& McMillen, 1994; Lewis et al., 1999; Mello & Broughman, 1996; Morton et al., 2008;
Seastrom, Gruber, Henke, McGrath, & Cohen, 2002) and Richard Ingersoll’s work (1999;
2002; 2003; 2005; 2006). Based on media coverage and commentary from influential
think tanks, the reporting about out-of-field teaching seemed to inject new worries into the
already dire image of American public school teachers. Out-of-field teaching, according to
these portraits, has been a persistent and widespread phenomenon in public schools since the
1980s and well into the 2000s in most academic subjects. More often, out-of-field teachers
are thought of as under-qualified teachers. Out-of-field teachers are automatically associated
with a negative influence on student learning, though the evidence for such claims has never
been clear.
Outside the United States, there are scattered statistics on out-of-field teaching. A 2006
UNESCO report showed that a large proportion of students in countries like Norway or
Canada were taught by out-of-field teachers in mathematics (UNESCO Institute for Statis-
tics, 2006, p. 65). An OECD report pointed out that lower qualification requirements in
subject matter are often used as a means to cope with teacher shortages in certain areas
(OECD, 2005). In countries like the Slovak Republic, it is estimated that “25% of pri-
mary classes, 30% of lower secondary classes, and 15% of vocational classes were taught by
teachers who did not have teaching qualifications, or were teaching out-of-field, or who had
already reached the official retirement age” (OECD, 2005, p. 49). Out-of-field teaching is
likely an international phenomenon but with insufficient internationally comparable statis-
tics. Currently there is no international database that tracks teachers’ out-of-field status

3
and experience across countries.
To understand the statistics that are available, we need to refer to a broader global
dialogue in which teacher policy is the key component for educational reform (Schleicher,
2011; Wilson et al., 2009). It is probably obvious that “the quality of an education system
cannot exceed the quality of its teachers” (Barber & Mourshed, 2007). Policymakers are
increasingly recognizing this view. In the recent International Summit on the Teaching
Profession 2011, the first of its kind, education ministers, teachers, and union leaders from
around the world convened to discuss the best practices to build a high-quality teaching force.
While teacher policy covers a whole spectrum of components, including teacher preparation,
recruitment, development, retention of teachers, and teacher evaluation and compensation,
out-of-field teaching connects to multiple aspects of these components. The study of out-
of-field teaching involves how teachers are trained in subject matter, how they are deployed
to schools, and how they are assigned to teaching posts. Out-of-field teaching also poses a
potential threat to the efforts that have been made to improve teacher quality. As Richard
Ingersoll put it, “highly qualified teachers may actually become highly unqualified if they
are assigned to teach subjects for which they have little background” (2003, p. 5). Rising
educational standards in curriculum are likely to require higher level mastery of subject
knowledge and innovative ways to teach. It is hard to imagine that teachers with little
training in the subject field are competent for such a task.
However, the research on out-of-field teaching has not yielded consensus among scholars.
Several questions remain inadequately answered, both internationally and within the United
States. To begin with, we do not fully understand the labor market mechanisms of supply
and demand that give rise to out-of-field teaching in different contexts. We are also unclear
about the prevalence of out-of-field teaching. Neither do we know out-of-field teachers’
demographics and educational background, working conditions, and professional status. Do
they work in systematically different schools than in-field teachers, and if so, what does their
work environment look like? Finally, is there any association between out-of-field teaching

4
and student learning?
This study aims to fill some of the gaps in the literature. I provide empirical evidence
of the scale and nature of out-of-field teaching in different countries. Specifically, I use
an international survey dataset with comprehensive and nationally representative data on
teachers to provide empirical evidence of the current status of out-of-field teaching in math
and science in public schools in 21 countries. This dataset, with a rich set of variables,
provides a rare glimpse into how teachers are trained and how they performed daily on
the job. The study aims to answer three main questions. First, are there cross-national
variations in out-of-field teaching? Second, what are the attributes and distribution of out-
of-field teachers across schools? Third, did out-of-field teachers engage in different teaching
practices and received differential professional support compared to in-field teachers? I adopt
a multi-level analysis approach to study the factors that correlate with the occurrence of out-
of-field teaching. Multi-level analysis allows me to investigate how individual, school, and
systemic-level factors play a part in shaping individual participation and school arrangements
in out-of-field teaching.
The policy implications of this study are many. The very definition of out-of-field teachers
invokes us to rethink the skills and knowledge that define today’s teacher. To understand
what constitutes an “out-of-field” teacher, one needs first to look into how teachers are trained
and licensed. Why are these practices legitimate and favored? Why are some skills more
important than others? From a school organizational point of view, out-of-field teaching
reflects how teachers are structured in schools around teaching activities. While some schools
are able to deploy teachers effectively to address learning needs in different subject areas,
many others often fail to do so. Why is this the case? To answer this question, one needs
to examine school-level characteristics related to teacher assignment and also regional and
national regulatory regimes that oversee school administration. The study of out-of-field
teaching could deepen our understanding of the operation of school organizations.
This study can inform policymakers about effective and innovative ways of regulating the

5
teacher labor market to ensure a sufficiently qualified teaching workforce and an accurate
match between teacher’s expertise in subject areas and student learning needs. These themes
have become increasingly important in today’s diverse classrooms across the world, with the
rising expectations for student achievement. Researchers and policymakers in education
realize that there is no silver bullet in policymaking that could be applied to each and every
situation. We need to equip them with a toolkit that they can apply in various settings.
Such an effort should be based on scientific and empirically grounded evidence and theories.
In the next chapter, I present a theoretical framework to understand out-of-field teaching.
Chapter 2 also reviews past studies on out-of-field teaching and their limitations. Chapter 3
introduces the data and research questions. Chapter 4 provides a descriptive analysis on
the attributes of out-of-field teachers. Chapter 5 includes multivariate regression analysis
comparing the relationship between the occurrence of out-of-field teaching at the school level
and systemic characteristics of each nation. I use a regression approach to compare teaching
practices between out-of-field and in-field teachers. In the last chapter, Chapter 6, I conclude
with providing reflections on the relationship between out-of-field teaching study and teacher
quality study.

6
 Chapter 2
CONCEPTUAL FRAMEWORK AND REVIEW OF RELEVANT
LITERATURE

In this chapter, I provide a conceptual framework to understand the policy concern of

out-of-field teaching. I also summarize the empirical evidence from the United States on the
origins of out-of-field teaching. In the following section, I first discuss the empirical measures
of out-of-field teaching and their applications. I then review the scale of out-of-field teaching
practice globally and current research on the effect of out-of-field teachers on student learning
outcomes. In the last section, I discuss the limitation of current literature.

2.1 Theoretical framework

2.1.1 Conceptualization

Out-of-field teaching is first discussed within the context of the United States. Though
this phenomenon has existed in American classrooms for at least fifty years, it was only
recently that out-of-field teaching has become a policy issue. Not too long ago, the require-
ment for a person to become a teacher was just having the basic ability in reading and
writing. Traditionally, some people believe that “pedagogical or methodological knowledge –
is of primary importance to be qualified. . . in-depth knowledge of a subject is less important
than in-depth skill at teaching” (Ingersoll, 2003, p. 10). Though the demand for teachers
to have a good understanding of subject matter knowledge came in the 19th century, not
many states required teachers to have subject matter knowledge beyond the secondary school
level (Sedlak, 2008). In the post-WWII era, more states started to mandate a minimum of
a four-year-college degree for new teachers. This movement aimed at professionalizing the
teaching profession in reaction to a widely circulated claim that American teachers lacked
basic content knowledge for the subjects they were teaching (Youngs & Grogan, 2010).

7
The professionalization movement gained new momentum after the No Child Left Behind
(NCLB) Act was enacted in 2002, in which subject matter proficiency is the key component
of the highly qualified teacher provision.3
The actual historical change in the definition of the teacher is beyond the scope of this
project. For the purpose of this study, it is important to note that the problematizing of the
out-of-field teacher is closely related to changing ideas on the skills and knowledge that define
a teacher. The critics of contemporary teacher education first questioned the subject matter
competency of elementary and secondary teachers (Conant, 1963; Koerner, 1963). In the
1980s, many of the critics and policymakers looked to reconstruct the identity of teachers
around subject matter knowledge and competency (Judge, Lemosse, Paine, & Sedlak, 1994;
Sedlak, 2008). The predominant view of the out-of-field teacher was thus from a knowledge-
deficit perspective. Increasingly, more educational reforms have put emphasis on the subject
matter training of teachers (Imig & Imig, 2008; Youngs & Grogan, 2010).
By definition, out-of-field teachers are those whose academic training does not match the
subject they are teaching. Out-of-field teachers lack sufficient training in the subject matter.
Thus they are deficient on two important aspects for being good teachers: subject matter
content knowledge and pedagogical content knowledge. Shulman once argued that “subject
matter knowledge and background in a content area affected the ways in which teachers
select and structure content for teaching, choose activities and assignments for students, and
use textbooks and other curriculum materials” (Shulman, 1988, p. 12). This stress on the
importance of subject matter knowledge has become the basic criterion to define today’s
teachers, as many high-quality teacher education programs put emphasis on chosen content
3 Subject matter proficiency is a key component in the highly qualified teacher provision in
NCLB. It requires a teacher to have all of the following: 1) a bachelor’s degree, 2) full state
certification or licensure, and 3) proof that they know each subject they teach. Teachers
must demonstrate competency by proving or obtaining one of these credentials: 1) a major
in the subject they teach, 2) credits equivalent to a major in the subject, 3) passage of a
state-developed test, 4) High, Objective, Uniform State Standard of Evaluation (HOUSSE),
5) an advanced certification from the state, 6) a graduate degree. For details, see U.S.
Department of Education (2002).

8
areas. But it is not just how competent a teacher is in the subject area that should be the
sole criterion to make a good teacher because “mere content knowledge is likely to be as
useless pedagogically as content-free skills” (Shulman, 1986, p. 8). Shulman further argued
that the content aspects of teaching entail knowledge about the subject and the way to
teach it. He defined subject matter content knowledge as “the amount and organization of
knowledge per se in the mind of the teacher,” and pedagogical content knowledge as “the
ways of representing and formulating the subject that makes it comprehensible to others”
(Shulman, 1986, p. 9). It is hard to argue that someone who does not have proper training in
the subject matter could develop the full ability to teach that subject without any additional
assistance. This argument is perhaps becomes the foremost reason to argue against out-of-
field teaching. Thus it is not uncommon to characterize out-of-field teachers as underqualified
teachers without basic competency (Ingersoll, 1999).
From an educational investment point of view, out-of-field teaching represents inefficien-
cies in the teacher labor market. Out-of-field teaching signifies the type of mismatch between
teachers’ subject matter knowledge and students’ learning needs that rendered the human
resources (i.e., skill sets and knowledge in the subject matter) accumulated throughout the
teacher training process under-utilized. Whatever the reasons might be for the cause of
out-of-field teaching, the fact that such practices exist shows the teacher market did not
channel the supply of particular teaching skills into classroom demands properly. In an
era where nations are investing in teacher training programs to prove educational quality,
the phenomenon of out-of-field teaching calls into questions whether such an investment is
yielding meaningful returns.
It should be noted that the lack of training in subject matter in itself may not undesirable.
If these teachers are given sufficient help at the starting stage, or continuing into their
service, they can still develop the essential skills and knowledge for teaching. This view
probably explains why in some education systems educational authorities recognize out-of-
field teaching as a temporary but legitimate practice in schools in order to cope with teacher

9
shortage problems. Teachers of this sort are required to develop sufficient subject matter
knowledge through a series of ad hoc training activities, and eventually to demonstrate
qualification in the subject matter (Eurydice, 2002). However, not many empirical studies
suggest that these policies are indeed being carried out.

2.1.2 The origins of out-of-field teaching

Apart from questioning the legitimacy of out-of-field teaching, another question has
emerged along with the research on out-of-field teaching: why does it happen on a large
scale in American classrooms? Out-of-field teaching does exist to a varying degree in many
other countries. But in the United States, the phenomenon seems especially persistent and
widespread.
Several explanations have surfaced over the years in discussions of the causes of out-
of-field teaching in the United States. The first argument came after the publication of
a report by the National Commission on Teaching and America’s Future (NCTAF hence
forward) in 1996, which claimed that out-of-field teaching, as a manifestation of the nation’s
poor teacher quality, was a result of inadequate teacher training. Several news commentaries
echoed this view in their syndicated columns.4 They claimed that teacher education in this
country is not rigorous enough to equip future teachers with sufficient knowledge for teaching.
The implication of this is that because teacher education in general is of poor quality, it
makes little difference for the employers to differentiate among people with different subject
matter background, thus leading to out-of-field teaching assignments. They proposed that
all prospective teachers go through a four-year undergraduate program and an additional
year of practice teaching to be fully qualified. Although such a view connects to the rhetoric
of ailing teaching quality in the United States, it fails to address why out-of-field teaching
4 Seefor example, To Help our Children, Invest in Teachers (David S. Broder, September
18, 1996, St. Petersburg Times, p. 10A). Education Insiders Protect Their Turf (Thomas
Sowell, September 23, 1996, St. Louis Post-Dispatch, p. 13B). A Need For Knowledge, Not
Skills (Maggie Gallagher, September 18, 1996, St. Louis Post-Dispatch, p. 7B).

10
continues to thrive in an era of increasing teacher qualification. Indeed, statistics show that
most teachers in the U.S. do at least have a bachelor’s degree, and a growing number of
them have a master’s degree (Ingersoll, 2003). The number of teachers with no certification
or license are dropping (Ingersoll, 1999). Findings also suggest that out-of-field teachers
typically have college training in a subject field (Ingersoll, 1999).
If the problem does not stem from teacher preparation (quality), could it be because of
quantity (e.g., the number of teachers produced)? The second view states that out-of-field
teaching is due to shortages of teachers in certain subject fields. This argument sounds
reasonable, but it also lacks compelling evidence.
In the United States, several factors have altered the supply and demand of teachers
historically including the fluctuation in population growth and shocks in the general labor
market such as those caused by the Great Depression (Sedlak, 1989). More recently, the
discussion on the supply of teachers has been focusing on the changing demographics and
state economy. From the supply side, as teachers of the baby-boomer generation step into
retirement age, more new teachers are needed to fill their positions (because the student
population is projected to keep growing). At the same time, many states that face stringent
economic constraints are proposing to cut the teacher workforce as a means to limit public
expenditures (Cavanagh, 2011). On the demand side, in several states lawmakers have
revamped the curriculum, which requires students to enroll in certain classes in order to
graduate from high school.5 These mandatory graduation requirements might also lead to
increase in demand for teachers in the short run, whether they have the qualification or not.
In addition, schools with tough working environments also use out-of-field teaching to cope
with teacher shortages. All these conditions could trigger out-of-field teaching.
Although each of these explanations all have possible connection with out-of-field teach-
ing, logical possibility does not necessarily mean causal relationship. Direct evidence sup-
porting these arguments is rather limited. There are also conflicting findings. Using School
5 Fora summary of state requirements, see:
http://www.cehd.umn.edu/nceo/topicareas/graduation/StatesGrad.htm

11
and Staffing Survey (SASS) data, Ingersoll (1999) found that some schools with no difficulty
in staffing reported a high level of out-of-field teaching. A closer look at the data also sug-
gests that out-of-field teaching existed in subjects that traditionally have had a surplus of
teachers, such as English and social studies (Ingersoll, 1999).6
Thus a direct link between teacher shortages and out-of-field teaching is not obvious.
Figure 2.1 illustrates a generic relationship.7 Out-of-field teacher hiring may be one of several
measures that could be used to cope with teacher shortage, but it does not necessarily have
a natural and necessary causal link with teacher shortage. Indeed, teacher shortage as a
phenomenon is also merely a product of many other forces that shape the teacher labor
market. From a policy perspective, to eliminate out-of-field teaching requires one to look
at the root of the phenomenon. In the increasingly complex labor market, using teacher
shortage as the explanation will only take us so far. A methodologically more plausible
question might be: What is the effect of teacher shortage on out-of-field teaching? In other
words, asking what is the effect of a specific cause, rather than asking what causes a given
effect (Holland, 1986).
Since the quality of teacher training or the quantity of teachers available are both inad-
equate to explain the phenomenon of out-of-field teaching, we turn to a third explanation.
This view is largely attributed to Richard Ingersoll in a series of studies (1999; 2002; 2003;
2005) in which he argued that the out-of-field teacher should be understood as an organiza-
tional issue (in the context of U.S. schools). The out-of-field assignment is not a phenomenon
solely because of emergencies, but is a common practice in school organizations. School ad-
ministration in the U.S. is fundamentally decentralized and highly autonomous. Much of the
decision-making regarding the staffing and hiring process is made within a school or district,
but there is a lack of direct inspection of the causes behind these decisions.8 Under this
6 Though less is known about whether the oversupply of teachers existed in certain subject
matters outside the U.S.
7 The figure is an illustration by the author based on study by Eurydice (2002); Gorard,
See, Smith, and White (2007); Ingersoll and Perda (2010).
8 Increasing inspection might raise administration costs.

12
 Out-of-field
Not enough
teaching
enrollment in
TE program

Increase class
size
High attrition
rate among Teacher Shortage
teachers
Rehire retired
teachers

Tough working
environment,
low Class
compensation Cancellation

Figure 2.1: Factors and policies that relate to teacher shortage

scenario, the school principal has large discretionary power in terms of assigning a teacher
to a teaching post. By contrast, teachers have little authority to decide their positions. In
many cases, one’s teaching assignment is made based on convenience or economic concern.
To match each teacher’s expertise with specific student learning needs is a time-consuming
endeavor. The flip side of this is that neither the school principal nor teachers will be re-
warded if they actually do so. Their compensation is not decided based on how well they
teach (Lazear, 2003). Burdened by other issues, principals might simply make decisions
based on their perception of a teacher’s ability to teach certain subjects or in some cases,
the decisions may be based simply on convenience. For many education administrators, al-
though subject matter competency is an important criterion when they make hiring decision,
more often they value other traits of a candidate: communicational skills, affection toward

13
children, or working ethics. Therefore the idea of using teachers with less qualifications in
certain subjects is legitimate on many occasions, especially in times of personnel absence
or shortage on a short-term basis. Such practices but may well evolve into more regular
practices if not addressed properly.
This view constitutes the basic lens this study used to examine out-of-field teaching.
Empirical evidence is limited on this issue. To date, most studies of out-of-field teaching
have been conducted in the United States. Arguably, if there are certain organizational
features pertaining to U.S. schools, a single-country study might not be ideal because of
limited variation on these features within the country. This study aimed to take advantage
of a multi-national dataset to compare diverse countries with one another.

2.2 Review of relevant literature

2.2.1 Measuring out-of-field teaching

To empirically measure whether a teacher has sufficient content knowledge training, past
studies typically followed one or several measures of out-of-field teaching suggested by the
National Center for Education Statistics (NCES) in the United States (Morton et al., 2008;
Seastrom et al., 2002):

• M-1 Out-of-field teaching by main assignments

• M-2 Out-of-field teaching by subject area taught

• M-3 Out-of-field teaching by class taught

• M-4 Number of students taught by teachers who are out-of field in a specific subject

All of these measures use one’s college major, minor, or subject field of certification as the
main indicator, but compare against different subjects. Out-of-field teaching status is defined
by two elements of teachers’ qualification: state certification/licensure and postsecondary

14
education. Though state certification traditionally is based on postsecondary course work
in the field to be taught, together with pedagogical coursework and teaching experience
(Seastrom et al., 2002), recently in an effort to attract more people into the profession some
states do allow individuals to obtain a state license as long as they take and pass the teacher
certification test.9 The literature suggests these following three indicators are used:

• I-1 Have at least a major or minor in the field of teaching

• I-2 Certified in the field

• I-3 Have a major or minor and is certified in the field

2.2.1.1 Out-of-field in main assignment

Using the three indicators, the main assignment measure is most straightforward. Each
individual teacher will be asked about their main teaching assignment, which will then be
compared to their educational credential indicators (i.e., I-1). The researcher could report
how many teachers are teaching in- or out-of-field (i.e., as defined by college major or mi-
nor) by their main assignments (Bobbitt & McMillen, 1994; Mello & Broughman, 1996;
National Commission on Teaching & America’s Future, 1996). There are two issues with
this measure. The first problem has to do with aggregation. In reality, individual teachers
could have multiple assignments within schools (sometimes even across schools). If one only
calculates the percent of out-of-field teaching using main assignments at the school level,
both the numerator and denominator are artificially smaller than they should be because
additional assignments are omitted. In survey data, information on the main assignment is
often self-reported. For individuals with more than one academic assignment, they may be
considered in-field by the main assignment but are teaching outside of their field of training
9 Theearliest case of such alternative route of certification is in the state of New Jersey.
See Copperman and Klagholz (1985) for reference.

15
in other assignments (Seastrom et al., 2002). Only considering main assignments would
underestimate the magnitude of out-of-field teaching.10
The second problem is that teachers are sometimes unclear about which assignment is
“main,” therefore omitting that additional assignment. In the Schools and Staffing Sur-
vey (SASS), a major database on teachers maintained by U.S. Department of Education,
many teachers reported their first assignment and “simply categorize the second as ‘other’ ”
(Seastrom et al., 2002, p. 21) thus creating a noisy measure of out-of-field teaching. An
example is special education teachers who provide individualized instruction to specific stu-
dents. Even when the surveyor collects all the assignment data from teachers, the “other”
category will not necessarily offer additional information on the out-of-field assignment sce-
nario.
Third, assignments are not equal at full time equivalence (FTE)11 levels. The same
math assignments could account for one’s full workload (therefore FTE=1), but to others
they might only account for half (FTE=0.5). The cases of whole class math teaching and
pull-out-individualized instruction are such examples.12 In this sense, assignment needs to
be weighted by FTE to reflect accurately the teaching load associated with each assignment.
But many empirical works hardly did that. Despite all the shortcomings, the advantage of
the main assignment measure is obvious: it is a straightforward measure and data are easy
to obtain.
10 The calculation will be: if Mr. A is certified in mathematics and teaches math for his
main assignment and at the same time teaches physics as a second assignment, then he is
an out-of-field physics teacher but not an out-of-field math teacher. The calculation of main
assignment measure for both math and physics will not be affected. But this measure will not
accurately reflect out-of-field teaching in physics because both numerator and denominator
are smaller by not counting multiple assignments.
11 FTE is a common way to measure and report teacher’s involvement in teaching or student
enrollment in school. An FTE of 1.0 means that the person being equivalent to a full-time
teacher/student. State administrative data typically reports each teacher’s FTE status.
12 Teachers who only offer individualized instruction are often considered as teaching on
part-time basis.

16
2.2.1.2 Out-of-field by subject area taught

The second measure (M-2) provides more detailed information of out-of-field teaching in
subject matter. The calculation is similar to the first measure but considers all teaching
assignments, separately by subject areas. Results using this measure usually report “among
all teachers teaching at least one science course, 20 percent are out-of-field” (Seastrom et al.,
2002, p. 22). The benefit of this measure is to allow the researcher to explore variations in
out-of-field teaching across subject fields. This takes into consideration multiple assignments
for a single teacher. The measure is constructed by subject area. A teacher who is teaching
both math and science but is only licensed in math would be counted as in-field in math
and out-of-field in science. Yet this measure also suffers from limitations similar to the first
measure, because teachers are not equal in their teaching loads; the researcher needs to weigh
each teacher by their full-time equivalence status. To account for multiple assignments across
subjects remains a challenge when using survey data such as SASS. Teachers with multiple
assignments need to be counted multiple times for different subject fields. If information on
additional assignments is not available, then bias is inevitable. Ingersoll (1999; 2003) uses
this measure. Like the first measure, the information is on the supply side, but “provides no
information on number of students exposed to such teaching” (Seastrom et al., 2002, p. 22).

2.2.1.3 Out-of-field by class taught

The third measure can be used to report the percentage of classes given in each subject
field that are taught by out-of-field teachers. This measure circumvents the challenge in
measuring multiple assignments. It can provide some information on students’ exposure to
out-of-field teaching. It also avoids the problem of different FTE status because the class is
used as the unit of calculation. Numbers of classes taught usually reflects the working load
of an individual teacher. But the information might not be precise since the class size could
vary by subject matter or grade level (especially at high school). In addition, this measure
requires the number of classes taught by an individual teacher, which is not always available.

17
2.2.1.4 Student taught in-field or out-of-field

The fourth measure is derived from the demand side. It eases the complication caused
by multiple assignments of teachers because the calculation no longer depends on teacher’s
feedback on assignment status. However it requires linkages between student and teacher
level data. It also requires more refined data on student attendance in specific classes. The
interpretation is also not easy. Researchers typically report a “percent of all students in each
subject area taught by teachers who are outside their areas of preparation” (Seastrom et al.,
2002, p. 23). SASS probably provides the most extensive measure of this kind (Bobbitt &
McMillen, 1994; Mello & Broughman, 1996; Morton et al., 2008; Seastrom et al., 2002).
The estimates based on the third and fourth measure require caution when interpreting.
Several NCES publications note that errors in estimation could be substantial. For example,
Morton et al. (2008) say that the “the standard error of this estimation is equal to 50 percent
or more of the estimates’ value” (pp.25, 27). The exact nature of these estimation errors is
not clear.
Practically, measures based on main assignments often give the lowest figure while mea-
sures based on classes taught give the highest estimates. The student-based estimates tend
to fall in-between. Examples of comparing different measures can be found in Seastrom et
al. (2002) tables B2, B4, B6, and B8.

2.2.2 Spread of out-of-field teaching

Most empirical work focuses on describing the level of out-of-field teaching. I start
this section by a comprehensive survey of the findings within the United States where this
literature is most well developed. International study on out-of-field teaching is emerging
and the evidence is limited. I briefly introduce current international findings later in the
section.
The NCTFA report first calculated the state-by-state percentages of the spread of out-
of-field teaching. The numbers ranged from 11 percent in Connecticut to 63 percent in

18
Alaska (National Commission on Teaching & America’s Future, 1996, p. 146).13 Using
SASS, Ingersoll (1999) later argued that over half of the teachers teaching physical science
in secondary schools (including chemistry, physics, earth science, and space science) did
not have an academic major or minor in any of these subject fields. About one-third of
secondary math teachers do not have teaching certificates in math. Over half of secondary
history teachers do not have an academic major or minor in history (Ingersoll, 1999). Several
NCES reports confirmed similar findings over the years (Bobbitt & McMillen, 1994; Lewis
et al., 1999; Mello & Broughman, 1996; Morton et al., 2008; Seastrom et al., 2002).
Table 2-1 summarizes published statistics from NCES using the main assignment measure
from 1987-2008. These statistics suggest that on average at the high school level, over time
more teacher assignments have been made based on the teacher’s college training. However,
in some subjects, such as mathematics, physical science, chemistry, and earth science, there
are still many out-of-field teachers.

13 According to Ballou and Podgursky (1996), the analysis also used SASS data.

19
Table 2.1: Percentage of U.S. public school teachers who taught high school without an undergraduate or graduate major in
their main assignments in selected subjects, 1987-88 to 2007-08

1987-88 1990-91 1993-94 1999-00 2003-04 2007-08

English 22.9 22.6 15.4 19 15.5 17.2
Foreign Lanauge na 46.6 30.5 37.5 na na
Mathematics 26.2 27.4 19.5 20.9 24 27.5
Science 20 20.1 15.8 16.6 12.8 16
Biology/ Life Science 33.1 41.8 29.4 34 18.8 23.9
Physical science 57.2 64.2 65.1 53.9 45.7 51.5
Chemistry 51.2 62.8 51.9 54.2 50.1 51.8
Earth science 69.3 69 65.2 64.5 60.2 66.8
Physics 63.5 66.9 51.3 51.1 42.8 42.3
Social Science 23.5 18.6 16.4 16.3 16.4 16.7
ESL 79.3 74.1 72 64 na na
Arts and music 11.1 7.9 5.5 7.9 8.4* 10.5*
Pysical Ed 12.1 8.4 7.1 11.1 na na
“*” indicates only includes art teachers
Source: Morton et al. (2008, p. 21); Schools and Staffing Survey (2010); Seastrom et al. (2002, p. 56)

20
 It is less clear about the characteristics of schools that are more likely to assign teachers
out-of-field. Past study suggests that out-of-field teachers are more likely to be concentrated
in high-poverty, small public schools (Ingersoll, 1999). Out-of-field teaching is more prevalent
in middle grades (7-8 grades) than in high school (Ingersoll, 2003). Minority and poor
students consistently receive fewer qualified teachers (Jerald & Ingersoll, 2002).
These studies all utilized SASS data and gave qualitatively similar findings. However,
findings from other data sources do not always fall into the same line. A report to the Califor-
nia state legislature reviewed subject assignment for more than 300,000 California teachers.
They found that the “misassignment” phenomenon, which is characterized as “the placement
of a certificated employee in a teaching services position for which the employee does not hold
a legally recognized certificate, credential, permit or waiver” (Credentialing and Certificated
Assignments Committee, 2008, p. 1), represented a 145.3 percent increase from 1992-95 to
2003-07. Yet misassignments only represent 6.33 percent of all teaching assignments. And
over half of the misassignments consist of teachers for English language learners (ELL). Us-
ing data from Michigan, Lynn and Keesler (2008) also reported no widespread mismatch
between teachers’ teaching assignments and fields of certification.14 Note that these studies
all utilized state administrative data in the post-NCLB era.
Using the National Education Longitudinal Study sample, Darling-Hammond, Berry, and
Thoreson (2001) found that 45 percent of math and science teachers taught a subject other
than their academic training. The authors argued that such a mismatch between teaching
and teacher education could simply reflect that some teachers are “teaching in a new field
while completing the requirements needed to be credentialed in the field”(p. 63). This finding
reminds the researcher to distinguish out-of-field teaching between new and veteran teachers.
It is reasonable that a new teacher may still be undertaking teacher training while teaching
in the classroom. However this argument still cannot explain why teacher training is not
14 Theout-of-field teaching measure used in this report only considered teacher’s main
assignment against area of certification, but did not take into account of multiple assignments
and FTE.

21
aligned with field of deployment in the first place on such a large scale.
At the state level, according to the newspaper Education Week’s annual Quality Counts
report in 2000, 11 states allowed teachers to teach outside of their certification areas for part
of their daily teaching without the need for special permission (Jerald & Boser, 2000).
Since the implementation of the No Child Left Behind Act in 2002, there has been a
steady increase in the number of state legislations that try to prevent out-of-field teaching.
As of 2000, 22 states developed a law to penalize schools or districts implementing out-of-
field teaching, yet there was barely any close monitoring or enforcement of the law. In the
same report eight years later, Education Week again pointed out that 32 states had at least
one policy, though not specified explicitly, to limit out-of-field teaching (Education Week,
2008). Eleven states now limit teacher’s out-of-field employment exceptions. Nine states
now require out-of-field teachers to earn additional alternative certification or accreditation.
Five states notify parents when out-of-field teaching takes place in schools. Four states
completely ban or cap the number of out-of-field teachers. It looks like out-of-field teaching
has gradually caught the attention of state policymakers.
Outside of the U.S., statistics on out-of-field teaching are rather limited. The UNESCO
report (UNESCO Institute for Statistics, 2006) gave two examples (Norway and Canada) of
high levels of out-of-field teachers. Reports from Trends in International Math and Science
Study (TIMSS) provide perhaps the most extensive information on teacher qualification.15
TIMSS reports reveal some interesting findings. For example, Mullis, Martin, Gonzalez, and
Chrostowski (2004) used TIMSS 2003 data and found that in many countries primary school
math teachers are not required to specialize in math. As a result, on average, 80 percent of
students are taught by teachers who majored in primary and elementary education. This
seems to be a general pattern in a majority of the TIMSS countries. Specifically, among
15 Because of the way TIMSS teacher survey is constructed, the teacher sample is not
nationally representative, but the student sample is. Any generalized reference to teacher
characteristics must be made in association with student population, which can be quite
cumbersome in the out-of-field teaching case.

22
4th -grade students, about one-fourth (26 percent) are taught by teachers specializing in
mathematics, 4 percent are taught by teachers specialized in or with a major in science, but
half (50 percent) of the students are taught by teachers who do not have any particular spe-
cialization (Mullis et al., 2004). The situation is similar for 4th -grade science teachers. On
average, science teachers for most 4th -grade students major in primary or elementary educa-
tion (80 percent). Martin and colleagues (2004) reported that “for primary education major,
about one-fourth (23 percent) of students were taught by teachers specialized in science,
7 percent in mathematics, and half (50 percent) not having any particular specialization”
(p. 253).
Similar findings are also found in the TIMSS 2007 report. Seventy two percent of 4th -
grade students are taught by math teachers whose major is in primary and elementary
education (Martin, Mullis, Foy, Olson, Erberber, et al., 2008). Seventy-two percent of 4th -
grade students are taught by science teachers whose major is in primary and elementary
education (Martin, Mullis, Foy, Olson, Preuschoff, et al., 2008). There is great variation
across countries. For instance, 94 percent of students in Italy are taught by teachers who do
not have a major either in elementary and primary education or mathematics and science,
compared to 15 percent in the United States.
Teacher preparation for 8th -grade math and science teachers is much different. On aver-
age, 70 percent of students enrolled in math classes were taught by teachers who specialized
in mathematics; 81 percent of students enrolled in science classes were taught by teachers
who specialized in biology, chemistry, physics or earth sciences (Martin, Mullis, Foy, Olson,
Erberber, et al., 2008; Martin, Mullis, Foy, Olson, Preuschoff, et al., 2008). It is more
likely for teachers who teach in higher grades to have a subject field specialization compared
to most elementary teachers, who are more likely to enroll in general education training.
The TIMSS data suggests that 27 percent of students enrolled in math classes at 8th -grade
and 19 percent of students enrolled in science classes at 8th -grade were taught by out-of-
field teachers. Table 2.2 summarizes the percentage of students taught by teachers who are

23
out-of-field.16
The estimates from TIMSS show great across-national variation. Yet the statistics are
not always consistent across years. For instance, math out-of-field estimates in Hong Kong
vary from 54 percent in 2003 to 40 percent in 2007. The estimates for science out-of-field
teaching is 22 percent in 2003 for Saudi Arabia and the it was 3 percent 4 years later
in 2007. Also because the statistics used in TIMSS are measured by the percentage of
students taught by out-of-field teachers, corresponding to the exposure measure mentioned
in Section 2.2.1.4, they are therefore not directly comparable to other national statistics,
which typically measure out-of-field teaching by the number of teachers.

16 According to the definition used by TIMSS publications. Out-of-field refers to subject

areas other than: general education, mathematics, education with specialization in math,
education with specialization in science, or science for mathematics teachers. For science
teachers, out-of-field major indicates major that other than education with specialization
in science, education with specialization in mathematics, biology/physics/chemistry/earth
science, or general education.

24
Table 2.2: Percentage of 8th -grade students in TIMSS whose teachers’ area of study is not
in math or science

2003 2007
COUNTRY
Mathematics Science Mathematics Science
Algeria na na 16 16
Armenia 25 13 40 42
Australia 42 39 39 30
Bahrain 7 13 7 12
Bosnia and Herzegovina na na 14 19
Belgium (Flemish) 24 35 na na
Botswana 28 25 21 16
Bulgaria 39 43 38 35
Chile 23 18 na na
Chiese Taipei 24 13 35 9
Colombia na na 29 16
Cyprus 10 12 11 10
Czech Republic na na 27 39
Egypt 7 13 12 7
El Salvador na na 41 47
Estonia 22 21 33 17
Georgia na na 17 8
Ghana 46 45 43 40
Hong Kong 54 25 40 30
Hungary 35 28 25 34
Indonesia 19 20 13 12
Iran 16 13 11 12
Israel 33 21 35 12
Italy 11 18 17 17
Japan 27 20 19 11
Jordan 7 9 18 16
Korea, Rep. of 5 7 4 6
Kuwait na na 2 9
Latvia 56 52 na na
Lebanon na 19 20 11

25
 Table 2.2: cont’d

2003 2007
COUNTRY
Mathematics Science Mathematics Science
Lithuania 15 28 11 19
Macedonia, Rep. of 1 6 na na
Malaysia 46 38 41 29
Malta na na 26 26
Moldova, Rep. of 23 19 na na
Morocco na 7 11 8
Netherlands 27 24 na na
New Zealands 53 31 na na
Norway 64 52 61 53
Oman na na 12 3
Palestinian Nat’l Auth. 5 13 4 9
Qatar na na 6 10
Philippines 14 22 na na
Romania 10 19 16 25
Russian Federation na na 17 20
Saudi Arabia 7 22 1 3
Scotland 37 15 22 14
Serbia 30 27 12 12
Singapore 44 25 50 31
Slovak Republic 29 35 22 na
Slovenia 19 22 28 16
South Africa 39 33 na na
Sweden 37 34 28 22
Syrian Arab Republic na na 11 15
Thailand na na 23 29
Tunisia 10 10 13 5
Turkey na na 9 9
Ukraine na na 12 12
United States 35 40 34 34
England 38 17 33 17
International Avg 27 24 22 19

26
 One European Union report gave qualitative information on the regulation each EU
country issues to allow/prohibit out-of-field teaching (Eurydice, 2002). According to that
report, out-of-field teaching, or by its definition “personnel who are not fully or appropriately
qualified” in subject matter, seems a widely used measure to cope with teacher shortages.17
Unlike the U.S., where out-of-field teaching is an inferior practice that causes much scrutiny,
most EU countries accept out-of-field teaching as an emergency measure that can be used
on a short-term basis. The exact nature of such acceptance of the out-of-field teacher is
not clear. The regulatory regimes seem well-developed, although no empirical evidence is
available on the scale of such practice. Neither do we know the degree to which those
regulatory policies are being implemented, or their effectiveness.

2.2.3 Effect of out-of-field teaching

Though the prevalent assumption about out-of-field teachers is that they must be less
effective than in-field counterparts in their ability to foster student learning (Ingersoll, 1999),
surprisingly there is only limited evidence regarding the effects of out-of-field teaching. In
an early study, Hawk, Coble, and Swanson (1985) used a small sample of 36 teachers and
found that out-of-field teachers know less mathematics than their in-field counterparts. Boe,
Shin, and Cook (2007) conducted a similar study using SASS data in 2007 and found that
out-of-field teachers are less well prepared in several teaching-related activities than in-field
teachers: ability to teach assigned subjects, select appropriate teaching materials, and plan
lessons effectively. Out-of-field teaching is also said to be associated with decreased teacher
morale and commitment (Ingersoll, 1999)
In terms of student achievement, Hawk, Coble and Swanson’s (1985) study using ANOVA
found that students perform relatively better when taught by in-field teachers in mathematics
17 More specifically, the following countries allow such practices with no specific restriction:
Belgium (Flemish and Germany), Denmark, France, Finland, Sweden, UK (England, Wales,
and Northern Ireland), Iceland, Norway, Bulgaria, and Lithuania. Estonia, Luxembourg,
Netherlands, Romania, Latvia, Slovenia, Czech, and Poland also allow such practice with
certain restrictions. For detailed information, see Eurydice (2002, p. 74)

27
classes. Rowan, Chiang, and Miller (1997) used a subsample of the National Education
Longitudinal Study of 1988 (NELS:88) and found a moderate effect of teachers who had
majored in mathematics in undergraduate or graduate degree on students’ mathematics
achievement. Their two-level hierarchical linear model suggests that compared to students
taught by teachers who did not have a major, students taught by teachers who majored
in math scored on average a 0.02 standard deviation higher. The size of this effect varied
depending on the average ability of students being taught in a school. Goldhaber and Brewer
(2000) used student-level data from the NELS: 88 survey. They found that students who
have been taught by teachers with no certification in the subject matter performed lower
than those whose teachers had either a standard, probationary, or emergency certificate.
Teachers who held a BA or MA degree in mathematics also positively influenced students’
math achievement. The effects were statistically insignificant, though, and the authors
admitted they were not able to disentangle the effect of out-of-field teaching from teacher
certification because they had no information on subject assignment. Using a comprehensive
dataset from Houston, Darling-Hammond, Holtzman, Gatlin, and Heilig (2005) found that
out-of-field teaching had a positive effect on TLI Math, TLI reading,18 SAT-9 Math, Aprenda
Math, and Aprenda Reading scores (5 out of 6 total student achievement indicators). That
study utilized student test data for several years and pooled data across time to conduct
OLS regression. This finding is rather surprising and counter-intuitive. However, as Ingersoll
suggested, out-of-field teachers might rely more on textbooks, and that kind of learning is
best captured by a standardized test (Ingersoll, 1999). He argued that one needs to look
beyond test scores to investigate teachers’ motivation and allocation of time to comprehend
the consequences of out-of-field teaching.
Another line of work is to use a direct measure of teacher’s subject matter knowledge
instead of using college major or license as a proxy. Hill, Rowan, and Ball (2005) designed
a specific test to measure teacher’s mathematical content knowledge. Their models show
18 TLI math and reading tests are assessments developed by the Learning Institute.

28
that teacher’s performance on this test is consistently positively related with student math
achievement gains in first grade and third grade. Interestingly, they also found that this
measure of teacher’s math content knowledge had little correlation with whether or not the
teacher was certified in math. Not surprising, using the same measure of teacher’s mathe-
matical knowledge, Hill, Ball, Blunk, Geoffney, and Rowan (2007) showed that those with
higher measurable math knowledge provided math instruction that was rich in representa-
tions, explanations, reasoning, and meaning.
To my knowledge, there is no study on the effects of out-of-field teachers conducted
outside of the United States.

2.2.4 Limitation of the literature

There are several limitations in previous studies. The data issue is perhaps the most
important. To date, there are only a handful of large-scale assessments of out-of-field teach-
ing. Data on teaching assignment and the educational background of individual teachers is
hard to obtain. Ingersoll suggested that data on out-of-field teaching could be a politically
sensitive issue for many local administrators at a time when more mandates are coming
from federal and state policy requiring locals to improve teacher quality (Ingersoll, 2003).
As mentioned before, in the United States most quantitative studies of out-of-field teaching
rely on information from SASS data. The benefit, of course, is that SASS is not collected
through local educational agencies but through teachers themselves. The teacher sample in
the SASS data is nationally representative, and the survey is repeated for multiple years.
Various researchers have used SASS data and others have reexamined their findings to as-
sure the validity of the findings (Bobbitt & McMillen, 1994; Lewis et al., 1999; Mello &
Broughman, 1996; Morton et al., 2008; Seastrom et al., 2002).
Yet even using a comprehensive dataset such as SASS could be limiting. For instance,
SASS did not require teachers to report all their instructional assignments. Teacher labor
force studies have pointed out that multiple assignments are a common phenomenon in

29
schools (Lynn & Keesler, 2008; Lynn et al., 2007). These multiple assignments sometimes
cover multiple subject fields and account for different amounts of workload (in terms of
FTE). Failure to report them could underestimate the magnitude of out-of-field teaching.
Though there are several competing theories that try to show the mechanism that drives
out-of-field teaching, there is little empirical evidence to support any of these theories. For
example, Ingersoll argued that out-of-field teaching is an organizational phenomenon. But it
is unclear to what extent school-level characteristics relate to the occurrence of out-of-field
teaching. It could be that out-of-field teaching is pertinent to the educational enterprise in
the United States. But a single country study is not likely to verify such a hypothesis simply
because there is not enough institutional variation within one country. Methodologically,
cross-sectional data such as SASS makes it difficult for researchers who try to quantify
the relationship between organizational changes at school levels and out-of-field teaching.
At the same time it is difficult to obtain quality indicators of out-of-field teaching and
school organizational factors such as the principal’s managerial style, teacher’s participation
in school management, and teacher supply and demand, which may partially explain why
studies of this sort are scarce. If these variables are systematically correlated with out-of-
field assignments but are not included as independent variables in estimating the likelihood
of individuals becoming out-of-field teachers, then the estimated coefficients will not be
consistent due to the omitted variable biases.
In addition, only a limited number of studies have evaluated whether out-of-field teachers
have an effect on student learning, and the findings are not conclusive (Boe et al., 2007;
Darling-Hammond et al., 2005; Goldhaber & Brewer, 2000; Rowan et al., 1997). There
are also methodological limitations in the previous studies. Hardly any of these studies
estimated the causal effects of students being taught by out-of-field teachers. Partly because
longitudinal data on students and teachers are not available in most cases, it is hard to
attribute the variation in student test score gains to teachers. In addition, a selection bias
for both teacher and students can cause serious estimation problems. The literature has

30
suggested that on the one hand, teachers are not randomly assigned to schools. Minority
teachers and teachers who have had less rigorous teacher training are more likely to move to
certain types of schools (Guarino, Santibanez, & Daley, 2006; Lankford, Loeb, & Wykoff,
2002). On the other hand, students and parents actively choose teachers as well (Lankford
et al., 2002). This “double sorting” will not be an issue if the characteristics of schools that
teachers and students choose to go to are independent of those characteristics associated with
out-of-field teaching. Yet empirical evidence suggests this might not be the case (Ingersoll
& Perda, 2010). Unfortunately no study to date has systematically taken this issue into
consideration.
Lastly, seldom have previous studies considered policy alternatives for out-of-field teach-
ing (including monitoring regulation). It is not clear what kind of regulation, school, or
national management regime could effectively curb out-of-field teaching. In the following
chapters I examine statistical evidence on how national and local policies can limit out-of-
field teaching from a comparative perspective. The advantage of cross-national studies is to
remind policymakers of the variety of practices in other countries. I will make policy recom-
mendation on how to limit out-of-field teaching based on the empirical findings I generate
in the next chapters.
To sum up, past studies of out-of-field teaching have mostly been done within the U.S.
context. Studies have found widespread and persistent out-of-field teaching phenomena
in U.S. classrooms. Internationally, there are emerging statistics reporting on out-of-field
teachers, but the full scale of such practices is not clearly understood. There are compet-
ing theories on the mechanisms that drive out-of-field teaching in schools. But no study has
systematically verified these arguments internationally. Lastly, the effect on student achieve-
ment of being taught by an out-of-field teacher is not known either. Guided by this literature
review, I discuss my research question and the data utilized to answer these questions in the
next chapter.

31
 Chapter 3
RESEARCH QUESTIONS AND DATA DESCRIPTION

In this chapter, I describe the data I used and the research questions that are to be
answered in the subsequent chapters. The research questions were derived from previous
studies in the United States.

3.1 Research questions

Q1. What is the distribution of out-of-field math and science teachers in the public schools
in the 21 study countries?

Q1 follows from the basic question: what is the prevalence of out-of-field practices in math
and science? I chose to focus on math and science teachers in this project for several reasons.
The first and foremost is the general agreement that a high-quality math and science teaching
force is essential for higher students’ performance in math and science (National Research
Council, 2002). The importance of developing math and science talents has been discussed
extensively by policymakers and researchers. They argue that people with high skills in these
fields form the backbone of the economy, which is driven by service and technology in the
21st century (Lowell & Salzman, 2007; National Academy of Sciences, National Academy
of Engineering, & Institute of Medicine, 2007). The obvious challenge toward achieving
this goal is how to maintain an adequate supply of math and science teachers, which has
proven to be difficult even for the developed nations (Ingersoll & Perda, 2010; Ladd, 2007;
National Comprehensive Center for Teacher Quality, 2009). Compared to teachers in other
fields, math and science teachers are facing unique labor market conditions. The demand for
people who have a background in math and science is particularly high in service and high-
end manufacturing sectors. Occupations outside of teaching that demand a great amount

32
of math and science skills offer higher wages. This in turn causes the numbers of math and
science teachers to be most susceptible to shortages compared to the numbers of teachers
in other subjects (Ladd, 2007). These findings support the view that a math and science
educator faces a different kind of supply and demand structure in the labor market, and that
dedicated analysis is required to understand the factors related to such differences.
Another reason for emphasizing math and science teachers is that compared to other
content areas, researchers have accumulated substantial knowledge on teacher preparation
and teacher effectiveness in these two subject areas. The accumulated knowledge not only
comes from the United States (Hill et al., 2005; Lee & Krapfl, 2002), but also internationally
(Schmidt et al., 2008, 2007). Yet despite the growing efforts to strengthen teacher education
at the higher education level globally (Luschei & Chudgar, 2011), we rarely ask whether
qualified candidates are in fact assigned to teach the subject that they are trained to teach.
Building on previous findings from the study of the teacher labor market, this study takes the
position that it is not enough to focus only on the supply side of the story; it is also important
to ensure that highly trained teachers are actually deployed to the positions that need them
most. Thus by asking this question, I intend to examine formally the distribution of teachers
in math and science, and the extent to which they do not have sufficient qualifications.
One distinction I make in this dissertation is that I focus only on public school teachers.
The rationale behind this decision is that the teacher labor markets facing public and pri-
vate schools are very different. There is compelling evidence showing that private schools
operate under different conditions than public schools. In terms of teacher-related issues,
public schools, typically enjoy a guaranteed supply of teachers from government-funded
teacher education institutions, whereas private schools, depending on their purposes, are
often recruiting from a different sector of the labor market. For example, Andrabi, Das,
and Khwaja (2008) described the hiring process of private schools in Pakistan, in which
the candidate pool is made up of largely young, single, moderately educated, and untrained
women from the local labor market, while government schools are able to recruit male and

33
educated teacher candidates because they are able to offer them higher wages and social
status. Similar processes have been found in studies of private schools teachers in many
other settings (McEwan & Carnoy, 2000; Muralidharan & Kremer, 2008; Psacharopoulos,
1987). Because the sustainability of private schools in many settings depends on the supply
of low-cost teachers, it is clearly not meaningful to treat the teachers who work in different
types of institutions as the same. Since I could not obtain more detailed information about
the operation and hiring practices of private schools in TALIS countries, I decide to focus
only on public school teachers. I should point out that in many developing nations where
universal secondary education has not been achieved, public schools usually enjoy certain
privileges compared to private schools in terms of filling teaching vacancies. Therefore the
estimates of out-of-field teaching using only public school data probably underestimate the
national out-of-field teaching in general.

Q2. What is the profile of out-of-field teachers within each of the study countries?
Q2.1 What is the demographic background of out-of-field teachers in each country?
Q2.2 What is the educational background of out-of-field teachers in each country?
Q2.3 What is the professional status of out-of-field teachers (e.g., opportunities for profes-
sional development; types of students they are teaching; whether or not they receive appraisal
for their work)?

After examining the overall out-of-field teaching practices addressed by Q1, I will study
the composition of this group of teachers and compare it to the majority of the teaching force
in each country. Popular perceptions about out-of-field teachers are that they are young
teachers who are assigned to subjects outside their training because there is no immediate
opening for the kind of jobs that match their expertise. They are just completing their train-
ing for teaching another subject on the job (Darling-Hammond et al., 2001). Interestingly,
there is not much empirical evidence to support this claim. There are scattered findings in

34
the United States that show that out-of-field teachers tend to be veteran, experienced teach-
ers (Ingersoll, 1999). Benefiting from the rich teacher background information provided in
TALIS, this question addresses the demographic backgrounds of out-of-field teachers. It is
also of interest to look at how demographic attributes of out-of-field teachers vary across na-
tions. Several labor market mechanisms are also worth considering. Are out-of-field teachers
used as a means of flexible labor deployment in schools (e.g., they are hired on a part-time
basis with less job security)? Are out-of-field teachers actually temporary teachers who only
cover the position in the short run? Do out-of-field teachers consist of a reserve of highly
educated but less-experienced teachers? Some countries allow out-of-field teaching as long as
the school provides teachers opportunities to acquire in-field status within a reasonable time.
Can out-of-field teachers get enough help on the job to develop subject matter knowledge
(e.g., professional development opportunities or appraisal from school administrators)?

Q3. At what kinds of schools do out-of-field teachers work?

Proposition 3.1 Out-of-field teachers are more likely to concentrate in schools that have
difficulty in recruiting teachers.
Proposition 3.2 Out-of-field teaching exists in schools where the school principal lacks
administrative leadership.
Proposition 3.3 Cross-nationally, the relative autonomy of hiring teachers at the school
level should interact with the above two factors and collectively shape the level of out-of-field
assignments.

Generally, there is no systematic information on the working environment of out-of-field

teachers outside the United States. We have limited knowledge regarding the hiring strategies
of schools that make out-of-field assignments. Ingersoll (1999) described some characteristics
associated with American schools that have larger numbers of out-of-field teachers (small and
private ones). Pertinent to my study, the labor market literature suggests that schools in

35
large cities and rural areas are susceptible to shortages in math and science (Ladd, 2007). In
large cities, alternative employment opportunities are likely to lure math and science talents
away from teaching while in rural areas, tough working conditions and low living standards
also offer less incentive for qualified math and science teachers. I empirically examined
whether the distribution of out-of-field teachers follow these previous findings.
The three propositions were formulated according to current understandings of the rela-
tionship between out-of-field assignments and school factors. Proposition 3.1 aimed to test
whether out-of-field teaching is associated with teacher shortages, which is widely believed to
be the main cause of such practices. Proposition 3.2 tested another hypothesis: out-of-field
assignment is due to the principal’s discretionary power in staffing. If this argument is true,
one would expect that out-of-field teaching occurs more often in schools where principals
solely decide the assignment of teaching positions without being held accountable; out-of-
field teaching is less prevalent in schools where daily management is collectively controlled
by teachers or through a centralized regime with high external accountability. Since these
arguments arise in the context of American public schools, which is also characterized by a
high degree of localized autonomy in hiring teachers with little external inspection (Ingersoll,
2006), one would expect that in the international context, the autonomy of hiring may in-
teract differently with the administrative leadership and teacher shortage factors in nations
where the institutional arrangements of hiring and assignment vary. This will be examined
in Proposition 3.3.

Q4. How do math and science out-of-field teachers in public schools differ from their in-field
colleagues?
Q4.1 Do out-of-field teachers spend different amounts of time on teaching practices than
in-field teachers?
Q4.2 Are out-of-field teachers equally likely to participate in professional activities in
schools?

36
 Q4.3 Do out-of-field teachers have more opportunities to develop their skills and knowl-
edge on the job?

Up until now, there has been limited empirical evidence regarding the differences in
teaching behavior and professional status between out-of-field and in-field teachers. Based
on the assumption that out-of-field teachers lack sufficient subject matter knowledge, and
consequentially pedagogical subject matter knowledge, out-of-field teachers may rely more
on textbook and routinized teaching practices (Ingersoll, 1999). Therefore, one would
expect out-of-field teachers to use a textbook-driven, more structured teaching approach in
the classroom, compared to the hands-on teaching that emphasizes student engagement and
use of different activities. In addition, if out-of-field teachers have poor training overall, we
could expect to see the amount of time they spend on school activities differ from in-field
teachers. For example, out-of-field teachers may spend more time controlling and managing
students than teaching if they cannot control the classroom properly. Another question
that is important to ask is whether out-of-field teachers are part of the school teaching
community. This question builds on the notion that out-of-field teaching is a transient
position in schools. Under such a scenario, we could expect out-of-field teachers to participate
less in collegiate collaboration or exchange of ideas. Lastly, I was interested to see whether
out-of-field teachers are offered on-the-job learning opportunities. If we presume that the
out-of-field teachers do not have the same amount of extensive training of subject matter
knowledge, one natural remedy is to develop such knowledge while they are on the job. So this
question particularly compared learning opportunities such as feedback about teaching from
the principals, participation in professional development activities, and receiving mentoring
from veteran teachers, etc.
Questions one through three were answered using descriptive methods. I compared mean
statistics and used graphs to present the findings. Questions two and three were also exam-
ined jointly using multivariate regression methods for each country. This technique ensured

37
that I consider multiple sources of influences at the same time. Question four was exam-
ined through a series of regression models that take into account teachers’ background and
school characteristics simultaneously. Detailed model specifications are presented in Chapter
5 where I show how I formally employed these methods.

3.2 Data

In this study, I rely on the data from OECD’s Teaching and Learning International Survey
2008 (TALIS 2008) data. TALIS is “the first international survey to focus on the working
conditions of teachers and the learning environment in schools” (OECD, 2009, p. 18). TALIS
is a cross-sectional survey database that includes lower secondary school- and teacher-level
data from 24 countries, over 4,000 schools, and over 70,000 teachers. Table 3.1 lists all the
countries participating in TALIS. European Union countries comprise the biggest proportion,
with a total of 19 countries.19 Two countries are from Asia (South Korea and Malaysia),
one from South America (Brazil), one from North America (Mexico), and one from Oceania
(Australia). Due to data quality issues, Iceland was not included in my analysis.20
One issue in international comparisons is the comparability problem. Since nation states
structure their education systems differently, it is important for researchers to make reason-
able comparisons. In TALIS, as shown in Table 3.2, the term “lower secondary” is used to
described different levels of schooling cross-nationally. Although the grade level varies from
country to country, it is generally believed that students began to receive departmentalized
instruction at similar grades as sampled in TALIS. As discussed previously, I focus on public
school teachers in the analysis, who make up the majority in TALIS sample.

19 Iceland and Turkey are not yet part of the EU, but they are official candidates. Norway
is not part of EU, though it is highly integrated into EU’s economy and regulation.
20 For reasons not disclosed to outside researchers, Iceland withdrew from the international
database (OECD, 2010). Therefore I decided to exclude it from my analysis.

38
Table 3.1: Participatant Countries in TALIS 2008

OECD Countries
Australia South Korea
Austria Mexico
Belgium (Flemish) Norway
Denmark Poland
Hungary Portugal
Iceland Slovak Republic
Ireland Spain
Italy Turkey
Netherlands
Partner Countries
Brazil Malaysia
Bulgaria Malta
Estonia Slovenia
Lithuania
Source: OECD (2010)

39
 Table 3.2: Description of samples

Country Grade sampled PSU MOS Total school Total teachers Public schools Total Math and
Unit sample sample sample Science teachers
Australia 7-10 S 149 2275 83 583
Austria 5-8 S 248 4265 209 989
Brazil 5-8 or 6/9 T 380 5834 257 1653
Bulgaria 5-8 T 199 3796 193 1036
Denmark 7-10 S 137 1722 76 537
Estonia 7-9 T 195 3154 189 730
Hungary 5-8 T 183 2934 163 728
Ireland 7-9 S 142 2227 55 516
Italy lower secondary T 298 5263 268 na
Korea 7-9 T 171 2970 110 738
Lithuania 5-10 T 206 3535 200 761
Malaysia 7-9 S 217 4248 208 987
Malta Form I - Form V NA 58 1142 30 265
Mexico 7-9 T 192 3368 149 930
Netherlands ISCED 2a S 39 484 33 114
Norway 8-10 T 156 2458 145 618
Poland 7-9 S 172 3184 158 794
Portugal ISCED 2 T 173 3046 130 857
Slovak Rep 5-9 T 186 3157 158 872
Slovenia 7-9 T 184 3069 177 843
Spain First - Fourth Compulsory T 193 3362 138 881
Secondary
Turkey 6-8 T 193 3224 140 707
Belgium (F) First stage of secondary education T 197 3473 83 667
a: ISCED 2 in the Netherlands consists of lower secondary education (first three years of pre-university education),
the first three years of senior secondary education and all four years of pre-vocational secondary edcuation
Source: OECD (2010)

40
 TALIS particularly focused on the role of teachers in the following aspects:

• The role and function of school leadership;

• How teachers’ work is appraised and the feedback they receive;

• Teachers’ professional development;

• Teachers’ beliefs and attitudes about teaching and their pedagogical practices.

3.2.1 Measuring out-of-field teaching

In the TALIS teacher questionnaire, teachers were first asked in Question 34 to identify
from the following list the subject they are teaching:

• Reading, writing and literature

• Mathematics

• Science

• Social Science

• Modern Foreign Language

• Technology

• Arts

• Physical education

• Religion

• Practical and vocational skills

• Others

41
 In this project I focus only on math and science teachers. Italy dropped out of the
analysis at this stage because the subject assignment variable was entirely missing for all
teachers. Teachers were then asked whether the teaching of this subject was part of one’s
academic training (Question 36). This was a yes/no type of question. I used the answer to
this question to construct my out-of-field teaching indicator. To focus on math and science
teachers who work in public schools, I first linked the teacher data and the school data for
each country, then used the public/private school variable provided in the school dataset to
restrict my dataset. By doing so, Ireland dropped out of my analysis due to missing linking
identification between teacher and school.
Using these variables, I constructed an out-of-field teaching measure. Using the method
discussed in Chapter 2, I calculated the percentage of out-of-field teachers (of the total teacher
population) in math and science area by schools (Equation 3.1). This measure became my
dependent variable in the subsequent analysis. In the analysis, I focused on math and science
teachers in particular. There are certainly limitations in the TALIS data. For instance, even
though it was common for teachers to take multiple assignments at lower secondary levels,
the teacher questionnaire does not differentiate between primary and secondary teaching
assignments. Therefore I was not able to discern whether out-of-field teachers are teaching
multiple subjects at the same time. I was also not able to differentiate out-of-field teaching
by assignment types. Secondly, TALIS does not provide a measure of FTE (full-time equiv-
alence), which has been shown in the U.S. literature to affect the estimation of out-of-field
teaching (Seastrom et al., 2002). I used a proxy measure of FTE to generate alternative
estimates of out-of-field teachers. These considerations made the information provided by
TALIS less than ideal. Despite these limitations, TALIS data make a unique contribution to
this research since this is the first time internationally comparable statistics on these issues
were made available to researchers.

T otal number of math and science OF T per school

P ercent OF T per school = (3.1)
T otal number of math and science teachers per school

42
 TALIS also provided a rich set of organizational variables at the school level. The strength
of the TALIS data was that it included many measures regarding teaching practices and
school environment, which were not available in most administrative data. These variables
included conventional school characteristics such as school size, locality, percentage of minor-
ity students, resource shortages, etc. TALIS included contextual variables such as whether
or not teachers could participate in hiring decisions; how much authority teachers had in
determining course content; and whether the school had a mentor teacher program. These
variables were used by the TALIS team to create several composite measures of school en-
vironment, which I used in my analysis. Exploring the relationship between these variables
and individual teachers’ out-of-field assignment expands our understanding of the influence
of school organization on teaching assignment. Table 3.3 describes the variables used in this
analysis:

43
 Table 3.3: Description of variables

Variable Type Description Source

Name
BTG01 categorical gender Teacher
BTG02 ordinal age Teacher
BTG03 ordinal part/full time Teacher
BTG06 categorical contract teacher Teacher
BTG07 ordinal educational background Teacher
BTG08A continuous time spent on teaching (hours) Teacher
BTG08B continuous time spent on preparing lesson (hours) Teacher
BTG08C continuous time spent on administrative duty (hours) Teacher
BTG09 ordinal years or experience Teacher
BTG11G1 categorical mentoring or peer observation Teacher
BTG12 continuous days spend on professional development Teacher
BTG30J ordinal observe other teacher’s class Teacher
BTG39B ordinal student ability Teacher
BTG41B continuous % of time spent on keeping order Teacher
BTG41C continuous % of time spent on teaching and learning Teacher
BCG08 categorical public/private School
BCG10 categorical school locality School
BCG12 continuous school size School
BCG15H ordinal principal informs teacher updating School
their skills and knowledge
TPRATIO continuous teacher-pupil ratio School
AVGCLSIZ continuous school mean class size School
PEDUATT3 continuous students with at least one parent/guardian who School
completed ISCED 3 or higher (school mean)
PEDUATT5 continuous students with at least one parent/guardian who School
completed ISCED 5 or higher (school mean)
NVREVAL categorical no appraisal or feedback received by the teacher Teacher
AUTHIRE continuous index of autonomy of hiring teachers School
LACKPERS continuous index of lack of personnel resources School
ADMINL continuous index of administrative leadership School
TSRELAT continuous index of teacher-student relationship Teacher
CCLIMATE continuous index of classroom climate Teacher
SELFEF continuous index of self efficacy Teacher
TBTRAD continuous index of direct transmission beliefs about instruction Teacher
TBCONS continuous index of constructivist beliefs about instruction Teacher
TPSTRUC continuous index of structuring practice Teacher
TPSTUD continuous index of student-oriented practices Teacher
TPACTIV continuous index of enhanced activities Teacher
TCCOLLAB continuous index of professional collaboration Teacher
TCEXCHAN continuous index of exchange and coordination for teaching Teacher

44
 Chapter 4
ATTRIBUTES OF OUT-OF-FIELD TEACHERS

In the past two decades, the study of out-of-field teachers has emerged as one of the
subfields of teacher quality study in the public policy circle in the United States. Researchers
have used comprehensive national data to examine the characteristics of out-of-field teachers
and the effect of being taught by out-of-field teachers on student achievement. The interest in
this topic builds on a shift in policy focus to require teachers to display competency in subject
matter knowledge. While a similar trend to use university-based teacher education as the
primary means to train future teachers has also taken place internationally, little is known
about whether such efforts actually improve teacher quality. In order to identify parallels
to the out-of-field teacher studies in the United States, in this chapter I use the TALIS
2008 data to describe the attributes of out-of-field teachers in 21 nations.21 The description
includes teachers’ demographic background, educational attainment, and working conditions.
Because the TALIS sample consists of in-service teachers and provides little information
about the teacher education background of these teachers, my analysis therefore primarily
focused on teachers who were already on the job. Whether or not out-of-field teachers
received different kinds of teacher education, and in what kinds of programs, are left for
future analysis. As discussed in Chapter 3, I chose to limit my sample to math and science
teachers who work in public schools in each country. Not only is the quality of math and
science teachers of central interest in many policy initiatives around the world, but also are
the labor market conditions faced by people who are trained in math and science fields that
differ from other disciplines. In addition, the choice of public school teachers over private
21 As previously sated, Iceland withdrew from international database after TALIS was
published. Ireland does not provide a link between school and teacher, and therefore excluded
in my analysis. Italy does not provide a subject matter identifier therefore I could not
distinguish subject matter teachers. I exclude Italy for this reason. These exclusions lead to
21 countries in final analysis.

45
school teachers is because public and private institutions’ recruiting patterns and potential
teacher candidate pools are different in several aspects. Due to limited information regarding
the operation of private schools in many countries, I chose to focus on public schools.

4.1 The distribution of out-of-field teachers across countries

Figure 4.1 shows a comparison of out-of-field teaching in math and science subjects among
public school teachers in 21 TALIS countries.

Poland
Hungary
Lithuania
Portugal
Slovenia
Bulgaria
Mexico
Korea
Slovak Rep
Malta
Netherlands
Austria
Spain
Belgium (F)
Norway
Australia
Turkey
Estonia
Malaysia
Denmark
Brazil

0 5 10 15

Figure 4.1: Percentage of out-of-field teachers in math and science, unweighted. For inter-
pretation of the references to color in this and all other figures, the reader is referred to the
electronic version of this dissertation

The level of out-of-field teaching ranged from 0.2 percent in Poland to 15.7 percent
in Brazil. The international mean of out-of-field teaching in math and science teachers
who teach in public school is 9.7 percent, which means one in ten teachers in math and

46
science was teaching out-of-field. Figure 4.1 clearly suggests that there is cross-national
variation. In Bulgaria, Hungary, Korea, Lithuania, Mexico, Poland, Portugal, and Slovenia,
out-of-field teachers were less than 5 percent of the entire math and science workforce. On
the other hand, Brazil, Denmark, and Malaysia had relatively large numbers of out-of-
field teachers. As mentioned in Chapter 2, several EU countries allow out-of-field teaching
with certain restrictions. These countries are Belgium (Flemish and German), Denmark,
France, Finland, Sweden, UK (England, Wales, and Northern Ireland), Iceland, Norway,
Bulgaria, Lithuania, Estonia, Luxembourg, Netherlands, Romania, Latvia, Slovenia, Czech,
and Poland, according to Eurydice (2002, p. 74). Several of these countries could be found
in the TALIS database. However it is difficult to tell whether these countries have a low
or high percentage of out-of-field teachers in math and science due to lax regulations. For
instance, though Poland allows out-of-field teaching, it has the lowest number of out-of-field
teachers among TALIS countries. On the other hand it seems Denmark is taking advantage
of its regulatory regime and deploys more out-of-field teachers into classrooms. Therefore
there is no direction correlation between policy regulations and actual levels of out-of-field
teaching across nations.
One limitation of this presentation in Figure 4.2 is that it does not take into account
a teacher’s full-time equivalence (FTE) status. For example, if out-of-field teachers are all
part-time teachers who only have 50 percent of employment time and we ignore this in the
calculation, the estimated percentage of out-of-field teachers will be inflated.22 Unfortunately
TALIS does not provide a detailed measure of FTE. Alternatively, I used one’s employment
time (part-/full-time status) as a proxy measure to reflect the individual teacher’s FTE
status. In the teacher questionnaire, every teacher answered whether they were full-time,
part-time (50-90 percent of full-time hour), or part-time (less than 50 percent of full-time
hours). I used this variable as an estimate of one’s FTE. If a teacher worked full-time, I
coded FTE as equal to 1. If one was employed at 50-90 percent of full-time status, I coded
22 A
numerical demonstration of the effect of FTE on out-of-field calculation is shown in
Chapter 2, Section 2.2.1.1.

47
FTE as 70 percent, the center of this range; if one was employed at less than 50 percent
of full-time status, I coded FTE as 25 percent. I then used this variable as a weight in the
calculation. Figure 4.2 shows the weighted estimates. It is not difficult to notice that there
is little difference between the estimates in Figure 4.1 and Figure 4.2. Notably the out-
of-field teaching level in math and science shrinks for Brazil: now Denmark is the country
with the highest level of out-of-field teachers. This suggests that many out-of-field teachers
in Brazil were indeed working on a part-time basis. We will see later in this chapter that
part-time teaching indeed characterizes out-of-field teachers in several countries. Since there
is little discrepancy between the weighted and unweighted, for the sake of simplicity, all of
the following analyses were done using the unweighted method.

Poland
Hungary
Lithuania
Portugal
Mexico
Slovenia
Bulgaria
Korea
Malta
Slovak Rep
Netherlands
Austria
Belgium (F)
Spain
Australia
Estonia
Norway
Turkey
Brazil
Malaysia
Denmark

0 5 10 15

Figure 4.2: Percentage of out-of-field teachers in math and science, weighted by part-time
status

In the current comparative work on teacher quality, many researchers use TIMSS as their

48
major source of data (Akiba & LeTendre, 2009; Akiba, LeTendre, & Scribner, 2007). The
appeal of TIMSS is obvious. As the largest-to-date cross-national comparative educational
database, it includes more than forty countries over a wide range of levels of educational
development. However, as some critics have pointed out, the TIMSS data is ill-suited to
the purpose of answering teacher-related policy questions because it is designed to include
nationally representative students instead of teachers (Rutkowski, Gonzalez, Joncas, & von
Davier, 2010). To put it differently, the teacher data included in TIMSS can only be
used to draw inferences about teachers’ experience by a nationally representative sample
of students rather than about a country’s teaching force in general because of the way
sampling is designed in TIMSS. To illustrate this point, I present a comparison of estimates
using TIMSS 2007 data to generate out-of-field statistics. Since I did not intend to make
any generalization, but instead compared the same estimates using two difference sources of
data, I only included countries that appear in TALIS 2008 as well as in TIMSS 2007, and I
used the United States for reference.
The comparison is shown in Figure 4.3. Since I only had 9 countries for comparison, it
was hard to draw any definitive conclusion. If we were to make a casual judgment, estimates
from TIMSS generally are higher than estimates from TALIS. In the case of Norway and
Australia, the TIMSS estimates are way higher than TALIS’s. It is not clear what causes
the disparity between the two sets of estimates. There are several possibilities. First, the
two studies used different measures of out-of-field teaching. The one used in TALIS is more
generic (because TALIS targeted teachers from various fields) and the one used in TIMSS was
specifically tailored to math and science education. Second, since the teacher population in
TALIS was nationally representative and the one in TIMSS was not, one needs to be cautious
about generalizability. Also the teacher sample from TALIS includes teachers from multiple
grades, and in TIMSS only 4th - and 8th - grade teachers were sampled. Third, the TIMSS
data does not always differentiate between public and private schools.23 Up to this point it
23 Countries are allowed to decide whether to include private schools in TIMSS. But there

49
is less clear which factor would cause the difference in estimates, and more technical work
surely is needed to probe into this issue. The bottom line is, policy researchers should be
advised not to use the TIMSS findings as the sole evidence against a nation’s teacher quality
(as is the case for Australia, Malaysia, and Norway).

Australia
Bulgaria
Hungary
Korea
Malaysia
Malta
Norway
Slovenia
Turkey
USA

0 20 40 60
TIMSS TALIS

Figure 4.3: Comparison of out-of-field estimates from TALIS 2008 and TIMSS 2007 data
(8th grade)

If we place confidence in the estimates of out-of-field teaching, one follow-up question

is how unique out-of-field teaching practices in math and science are, compared to other
subject matters. In Figure 4.4 I provide a brief comparison among public school teachers
from different subject fields. I categorized teachers by subject matter into three categories:
math and science, humanities, and other.24 The degree of deployment of out-of-field teach-
is no indicator for research to differentiate between these two types of schools. Since TALIS
also contains private school teachers, I conducted a separate calculation between TALIS and
TIMSS using the full TALIS sample; the result barely changed.
24 The categorization followed OECD’s method (OECD, 2009, p. 100). Humanities include

50
ers in math and science was close to those in humanities in the majority of countries.25 In
general, out-of-field teaching was more prevalent in “other” subject matters. This finding is
not surprising because typically current university-based teacher training is more than likely
to focus on academic subjects. Regulations on teacher assignments are more common in
academic subject areas in many countries, but little evidence can be found that the same
regulations apply to teachers in non-academic subject matters, such as vocational education
and technology. Teacher qualifications in non-academic subjects tend not to include a spe-
cialized degree in these subject matters. This is the indication that more effort has been put
into providing professional training for teachers in math and science fields so that out-of-field
teaching can be kept under control.
Compared to the U.S. data that NCES compiled in the 1990s, the countries surveyed in
TALIS had substantially lower levels of out-of-field teaching. For example, NCES’s estimates
show that in the United States, 64.7 percent of middle school math teachers do not have a
major in mathematics; in science, the estimate is 49.6 percent. These numbers are higher
than Brazil, the country with the highest math and science OFT in the TALIS sample.
Although the U.S. data was calculated using earlier statistics, and no up-to-date nationally
representative statistics exist, the substantial gap in out-of-field teaching between the U.S.
and the rest of the world indicates that the U.S. is clearly not on the better side of the
teacher quality spectrum. Even when we consider the estimates from TIMSS, which probably
contained errors to a degree, the out-of-field level in math and science in the U.S. was about
30 percent, still substantially higher than majority of the TALIS countries. Of course there
is reason to believe that, with the enforcement of the highly qualified teacher provision in the
federal No Child Left Behind Act, states have geared up to reduce out-of-field teaching, yet
the sheer differences suggest tremendous efforts still need to be made in the United States
reading, writing, and literature, social science, modern foreign language, arts, and religion.
Other includes: physical education, vocational education, technology, and other unspecified
subjects.
25 Malaysia, Mexico, and the Slovak Republic might be exceptions.

51
 Australia Poland
Austria
Brazil Portugal
Bulgaria
Slovak Rep
Denmark
Estonia Slovenia
Hungary
Spain
Korea
Lithuania Turkey
Malaysia
Malta Belgium (F)
Mexico
0 5 10 15 20 25
Netherlands
Humanities
Norway
Math&Sci
0 10 20 30 Others

Figure 4.4: Percentage of out-of-field teachers by subject fields

in order to catch up with the rest of the world. Another caution that needs to be made
when interpreting the statistics is that I focused on public school teachers in the analysis.
As stated in Chapter 3, the estimates based on public school teachers are likely to understate
the scale of out-of-field teaching.

4.2 Demographic background of out-of-field teachers

4.2.1 Gender

Figure 4.5 contrasts the gender composition between out-of-field and in-field teachers
across nations.26 A general observation that can be made from Figure 4.5 is that out-of-field
26 Countries are sorted in descending order by male teachers as percentage of out-of-field
teacher.

52
teachers’ gender composition is close to their in-field colleagues’ gender across countries.
A simple statistical test confirms our observation: the difference in gender composition is
statistically zero in 17 countries. In most countries, teaching is a feminized occupation, only
a handful countries have a male-dominant teaching force (male teachers are over 50 percent),
such as the Netherlands, Turkey, Denmark, Norway, Mexico, Australia, and Malta. In these
countries, there tends to be more males who are out-of-field teachers. Gender compositions
between out-of-field and in-field teachers were statistically different in 4 countries: Estonia,
Hungary, Portugal, and Turkey, where out-of-field teachers were significantly more male
than female. I should remind readers that this finding is derived from the math and science
teacher population. It remains to be examined whether the same observation applies to
other segments of the teacher labor force.

Netherlands
Turkey
Hungary
Denmark
Norway
Portugal
Mexico
Belgium(F)
Spain
Australia
Brazil
Austria
Malaysia
Estonia
Malta
Slovak Rep
Slovenia
Korea
Lithuania
Bulgaria
Poland

100 75 50 25 0 25 50 75 100

IFT OFT

Figure 4.5: Proportion of male teachers by out-of-field status.

53
 Male out-of-field teachers in several cases look quite different from female out-of-field
teachers. In Table 4.1, for example, in Denmark male out-of-field teachers are predom-
inantly older than female out-of-field teachers. In Turkey, male out-of-field teachers are
predominantly younger than female out-of-field teachers. In Korea, male out-of-field teach-
ers tend to be middle-aged, while female out-of-field teachers are distributed more evenly
across all age groups. Does this suggest that there is a gender preference in terms of choosing
out-of-field teaching as a job? Or does it reflect different opportunity structure caused by
gender role in the labor market?
I further broke down age distribution by both gender and out-of-field status. As described
in the three cases above, gender groups have different age characteristics in several countries.
Yet we should put such dissimilarities into a broader context to compare them with the male
and female distributions of in-field teachers. In Table 4.1 and Table 4.2, I calculated similar
statistics for in-field teachers for comparison; I found that the age distributions are similar
for male out-of-field and male in-field teachers. This means that gender composition among
out-of-field teachers reflects occupation-wide gender difference.
Overall, gender composition has more between-country variation than within-country
variation between out-of-field and in-field teachers. Such a pattern suggests that national
contexts have clearly influenced the characteristics of the mathematics and science teacher
labor market.

54
 Table 4.1: Age distribution of male and female OFT

Male OFT Female OFT

Country
<30 31-49 >50 <30 31-49 >50
Australia 21% 45% 34% 29% 57% 14%
Austria 0% 39% 61% 14% 59% 27%
Brazil 30% 53% 16% 20% 65% 15%
Bulgaria 0% 100% 0% 0% 32% 68%
Denmark 1% 51% 48% 8% 82% 9%
Estonia 31% 45% 24% 17% 47% 36%
Hungary 43% 35% 22% 0% 67% 33%
Korea 0% 86% 14% 23% 52% 25%
Lithuania 15% 22% 63% 0% 68% 32%
Malaysia 12% 85% 3% 22% 73% 5%
Malta 0% 100% 0% 17% 67% 17%
Mexico 4% 71% 25% 40% 50% 10%
Nethlands 14% 43% 43% 0% 0% 100%
Norway 5% 39% 56% 3% 51% 46%
Poland na na na 0% 100% 0%
Portugal 17% 63% 21% 7% 86% 7%
Slovak Rep 0% 85% 15% 33% 52% 15%
Slovenia 0% 86% 14% 2% 98% 0%
Spain 0% 88% 12% 15% 64% 21%
Turkey 40% 48% 12% 2% 98% 0%
Belgium (F) 57% 37% 6% 48% 19% 33%

55
 Table 4.2: Age distribution of male and female IFT

Male IFT Female IFT

Country
<30 31-49 >50 <30 31-49 >50
Austria 6% 42% 52% 5% 53% 41%
Brazil 29% 56% 15% 19% 72% 10%
Bulgaria 9% 44% 47% 5% 59% 36%
Denmark 5% 44% 51% 12% 59% 29%
Estonia 11% 47% 42% 9% 44% 46%
Hungary 8% 55% 37% 4% 55% 41%
Korea 8% 74% 18% 21% 76% 3%
Lithuania 1% 44% 55% 5% 52% 43%
Malaysia 16% 70% 14% 28% 65% 7%
Malta 23% 52% 25% 35% 52% 13%
Mexico 7% 67% 25% 14% 67% 19%
Nethlands 23% 27% 50% 17% 57% 26%
Norway 7% 50% 43% 14% 52% 34%
Poland 22% 53% 26% 8% 73% 19%
Portugal 8% 80% 12% 10% 81% 9%
Slovak Rep 8% 46% 46% 13% 42% 45%
Slovenia 13% 60% 27% 12% 66% 22%
Spain 5% 46% 48% 5% 64% 30%
Turkey 50% 33% 17% 51% 47% 2%
Belgium (F) 31% 46% 23% 22% 49% 29%

56
4.2.2 Age

In terms of age composition, the differences between out-of-field teachers and in-field
teachers are typically spotted at the two ends of the distribution. This suggests that out-
of-field assignments are made for two distinctive subpopulations of teachers: those who are
young or older27 as seen in Figure 4.6 and Figure 4.7. The dots in each line represents the
percentage of teachers. The blue dots indicate percentage of out-of-field teachers. The red
dots indicate percentage of in-field teachers (in math and science), who are the majority in
the teacher labor force. In Figure 4.6, for instance in Belgium (Flemish), fifty percent of
out-of-field teachers are under age 30 and only 26 percent of in-field teachers are under age
30.
At the lower end of the spectrum, nine countries have a relatively higher proportion (>20
percent) of younger out-of-field teachers. They are Belgium (Flemish), Australia, Hungary,
the Slovak Republic, Turkey, Brazil, Estonia, Mexico, and Korea. Belgium (Flemish), Aus-
tralia, and Hungary’s out-of-field teaching force consists of more than 30 percent of young
teachers who are below the age of 30. One immediate question is whether these young
out-of-field teachers are less qualified. Judging by their education attainment (Table 4.3),
this does not seem to be the case. Most of them have at least a bachelor’s degree. Is it
possible that these young teachers are at some sort of practice teaching stage before they
are eventually assigned to a particular subject category? Without an explicit variable that
measures their student teaching status in the questionnaire, I only have indirect evidence
about this argument.
In Table 4.4, I break down the years of teaching experience by age categories. It is
clear that there are countries like Australia where all young out-of-field teachers are first-
year teachers. Yet in a majority of countries, not all young teachers are novice teachers.
It is common for these young teachers to have 3 years of more teaching experience, which
27 For
the ease of creating graphs for comparison, I re-categorize the age variable in the
TALIS database into three categories: below 30, 30 to 50, and above 50.

57
 Bulgaria
Poland
Slovenia
Lithuania
Norway
Denmark
Spain
Portugal
Malta
Austria
Netherlands
Malaysia
Korea
Mexico
Estonia
Brazil
Turkey
Slovak Rep
Hungary
Australia
Belgium (F)

0 10 20 30 40 50
OFT IFT

Figure 4.6: Percentage of teachers under age 30 by out-of-field status

rules out the possibility that they are indeed student teachers. At the upper end of the
age distribution, not surprisingly, a majority of older out-of-field teachers also tend to be
veteran teachers. There are spotty cases such as Australia, Austria, and Bulgaria where a
small proportion of older out-of-field teachers also have little experience. These people are
likely to treat teaching as a second career, but in general such practices are not common
among TALIS countries. The data suggests that out-of-field teaching at least is not used as
as a second career for old-yet-inexperienced teachers.
Previous literature has suggested that within schools (Ingersoll, 2006), veteran teachers
tend to use their seniority to exert influence over teaching assignments. Following this
argument, one would expect to see a smaller proportion of older and experienced out-of-
field teachers and a larger proportion of younger and inexperienced out-of-field teachers.
Figure 4.7 seemingly supports this argument because at the upper end of the age distribution,

58
 Poland
Slovenia
Malaysia
Turkey
Slovak Rep
Brazil
Spain
Malta
Mexico
Portugal
Australia
Belgium (F)
Hungary
Korea
Estonia
Lithuania
Denmark
Austria
Netherlands
Norway
Bulgaria

0 20 40 60
OFT IFT

Figure 4.7: Percentage of teachers over age 50 by out-of-field status

there are clearly fewer out-of-field teachers compared to in-field teachers. At the lower end
of the age distribution (Figure 4.6), at least in the case of Belgium (Flemish), Australia,
Slovak Republic, Mexico, Estonia, and Hungary, many more young teachers are chosen to
teach out-of-field. The hypothesis seems to hold in these countries.
However, this theory does not explain the deployment of older out-of-field teachers. Such
a practice is evident in at least three countries: Bulgaria, Norway, and the Netherlands. A
closer look suggests that these teachers are also experienced, and highly educated. Presum-
ably they are not at the bottom of the school power structure. One explanation is that they
are not actually part of the community. These teachers are typically hired on a part-time
or short-term basis. In other words, they are not regarded as the “regular” teaching staff in
the particular; instead, they are like substitute teachers who answer to short-term requests
to teach something for which they do not have training, albeit they are experienced in other

59
 Table 4.3: Distribution of educational attainment of out-of-field teachers under age 30

Educational attainment
Country
Below BA BA MA or above
Australia 0% 100% 0%
Austria 93% 0% 7%
Brazil 19% 80% 1%
Bulgaria 45% 2% 54%
Denmark 0% 100% 0%
Estonia 30% 60% 11%
Hungary 0% 100% 0%
Korea 0% 100% 0%
Lithuania 100% 0% 0%
Malaysia 0% 97% 3%
Malta 0% 100% 0%
Mexico 0% 82% 18%
Netherlands 0% 0% 100%
Norway 29% 71% 0%
Portugal 0% 100% 0%
Slovak Rep 0% 0% 100%
Slovenia 0% 100% 0%
Spain 0% 0% 100%
Turkey 0% 100% 0%
Belgium (Flemish) 48% 30% 22%

aspects. This answer could only apply to a small proportion of out-of-field teachers in these
three countries; however the majority of out-of-field teachers have full-time employment and
hold permanent teaching positions (Table 4.5). It could be argued that this might be the
case because of their long-term general teaching expertise so that they are covering a class
temporarily. Qualitative information might help to confirm these hypotheses. These three
countries all allow out-of-field teaching as an emergency measure in case of teacher shortages
(Eurydice, 2002).28
Despite special cases of Bulgaria, Norway, and the Netherlands, the pattern of age dis-
tribution is quite consistent at this end of the spectrum: out-of-field teachers are less likely
28 More specifically,in Bulgaria and Norway, policies allow deployment of out-of-field teach-
ers without special restrictions; in the Netherlands, out-of-field teaching has to be in “related
subjects only” (Eurydice, 2002, p. 74).

60
to be old teachers in a majority of countries. The blue dots in Figure 4.7 always lie on the
left of the red dots except for Brazil.

61
 Table 4.4: Distribution of teacher experience by two age groups among out-of-field teachers

Age <30 years old Age >50 years old

Country
0-1 year 1-2 year 3-5 year 6-10 year 0-1 year 1-2 year 3-5 year 6-10 year
Australia 100% 0% 0% 0% 25% 0% 0% 0%
Austria 20% 48% 32% 0% 0% 0% 25% 0%
Brazil 11% 14% 22% 53% 0% 0% 0% 0%
Bulgaria 25% 17% 34% 23% 0% 0% 0% 13%
Denmark 8% 0% 4% 15% 0% 0% 0% 0%
Estonia 0% 73% 27% 0% 0% 0% 6% 5%
Hungary 32% 23% 45% 0% 0% 0% 8% 4%
Korea 0% 0% 61% 39% 0% 0% 0% 0%
Lithuania 0% 19% 61% 20% 0% 0% 0% 0%
Malaysia 0% 0% 100% 0% 0% 0% 0% 26%
Malta 22% 25% 46% 6% 0% 0% 0% 0%
Mexico 0% 0% 0% 100% 0% 0% 0% 0%
Netherlands 25% 19% 46% 10% 0% 0% 0% 0%
Norway 0% 24% 51% 25% 0% 0% 0% 0%
Poland 41% 31% 28% 0% 0% 0% 0% 0%
Portugal 0% 0% 0% 0% 0% 0% 0% 0%
Slovak Rep 53% 47% 0% 0% 0% 0% 0% 0%
Slovenia 0% 62% 3% 36% 0% 0% 0% 14%
Spain 0% 0% 0% 100% 0% 0% 0% 0%
Turkey 21% 54% 24% 0% 0% 0% 0% 0%
Belgium (F) 0% 28% 67% 5% 0% 0% 0% 0%

62
 Table 4.5: Contract type and part-time status of out-of-field teachers over age 50

Contract type Part-time status

Country
Permanent posi- fixed-term more fixed-term less Full-time Part-time
tion than 1 year than 1 year
Australia 78% 0% 22% 81% 19%
Austria 100% 0% 0% 88% 12%
Brazil 90% 0% 10% 47% 53%
Bulgaria 85% 2% 13% 100% 0%
Denmark 100% 0% 0% 100% 0%
Estonia 77% 6% 17% 64% 36%
Hungary 100% 0% 0% 100% 0%
Korea 100% 0% 0% 100% 0%
Lithuania 90% 6% 4% 50% 50%
Malaysia 100% 0% 0% 100% 0%
Malta 100% 0% 0% 100% 0%
Mexico 90% 0% 10% 59% 41%
Netherlands 100% 0% 0% 50% 50%
Norway 100% 0% 0% 88% 12%
Poland na na na na na
Portugal 100% 0% 0% 100% 0%
Slovak Rep 100% 0% 0% 100% 0%
Slovenia 100% 0% 0% 100% 0%
Spain 100% 0% 0% 100% 0%
Turkey 100% 0% 0% 54% 46%
Belgium (F) 100% 0% 0% 73% 27%

4.2.3 Employment status

In Figure 4.8 I show the proportion of teachers by their part-time/full-time employment

status for in-field and out-of-field teachers. Employment status is generally quite close be-
tween out-of-field and in-field teachers within each country. Graphically one could argue
that there are slightly fewer out-of-field teachers who work on a full-time basis. There is no
statistical difference when I formally test the difference between these two groups of teach-
ers in most of the countries. Notably there are six countries that almost entirely prohibit
part-time work among teachers. They are Bulgaria, Hungary, Korea, Malta, Poland, and

63
the Slovak Republic.
To compare between countries, I first separated two different types of part-time teachers
into a group labeled “part-time teachers.” Then I subdivided the sample countries into two
groups: those with larger than 20 percent of its entire teaching force in math and science
working part-time, and those with smaller than 20 percent of math and science teachers
working part-time (see Figure 4.9). Among countries with a high percentage of part-time
teachers (those to the right of the vertical line), we also witness even more out-of-field
teachers who work part-time. The differences in percentage of part-time teaching between
out-of-field and in-field teachers are typically between 5-15 percent in Mexico, Brazil, Turkey,
Estonia, and Lithuania.29 This suggests that when part-time working is institutionalized in
the education system, out-of-field teachers exacerbate such conditions by a large proportion.
The Netherlands is the only exception in this category where the proportions of part-time
teachers are virtually the same between out-of-field and in-field teachers.
Overall, out-of-field teachers are less likely to be hired on a full-time basis in the majority
of the countries.

29 Only in Estonia the difference is statistically significant at 0.05 level.

64
 Australia IFT IFT
OFT Netherlands
OFT
Austria IFT
OFT IFT
Norway
OFT
Brazil IFT
OFT IFT
Poland
IFT OFT
Bulgaria OFT
IFT
IFT Portugal
Denmark OFT
OFT
IFT IFT
Estonia Slovak Rep
OFT OFT
Hungary IFT IFT
OFT Slovenia
OFT
Korea IFT
OFT IFT
Spain
IFT OFT
Lithuania OFT IFT
Turkey
Malaysia IFT OFT
OFT
IFT
IFT Belgium (F)
Malta OFT OFT

Mexico IFT 0 20 40 60 80 100

OFT
IFT full-time
Netherlands OFT 50-90% time
0 20 40 60 80 100 <50% time

Figure 4.8: Distribution of teachers by part-time status

65
 Bulgaria
Hungary
Korea
Malta
Poland
Slovak Rep
Malaysia
Denmark
Portugal
Spain
Australia
Norway
Austria
Slovenia
Belgium (F)
Estonia
Turkey
Netherlands
Lithuania
Brazil
Mexico

0 20 40 60
OFT IFT

Figure 4.9: Percentage of teachers who work part-time by out-of-field status

66
4.2.4 Contract length

Another way to measure the working status of out-of-field teaching is to compare the type
of contract they signed with the schools. The TALIS database distinguishes three types of
contracts: fixed-term, for less than one school year; fixed-term, for more than one school year;
and permanent positions. Here I will focus on the first and last types of contract, because
the first type signifies a flexible labor deployment strategy and the last type represents the
norm in teacher hiring.
In general, in-field teachers tend to have longer and more stable contracts. This statement
reflects the smaller number of fixed-term contracts that are signed by in-field teachers, and
the larger number of permanent contracts among in-field teachers. In 18 of the 21 countries
in this analysis, less than 10 percent of in-field teachers had fixed-term contracts. The
exceptions were Poland, Portugal, and Brazil. In Figure 4.10, we see that out-of-field teachers
have a higher percentage of short-term contracts compared to in-field teachers in math and
science fields. Graphically, this observation reflects that the blue dots, which represent out-
of-field teachers, are in general to the right (therefore indicating higher percentage) of the red
dots, which represent in-field teachers. One extreme case is Belgium (Flemish), where close
to 50 percent of the out-of-field teachers were hired on a fixed-term short contract. Though
the Belgium case is not universal, five countries have larger disparities in the proportion of
teachers who signed fixed-term contracts between out-of-field and in-field teachers. They are
Hungary, Australia, Spain, Estonia, and Denmark.30 This finding indicates that out-of-field
teaching deployment tends to be more flexible than in-field in these countries. In Spain and
Brazil, though the short-term contract is common among out-of-field teachers, it does not
distinguish them from their in-field counterparts.
At the other end, in-field teachers are easily distinguished from out-of-field colleagues by
the sheer number of permanent contracts. The pattern is clear in 14 countries. In Figure 4.11,
in the Slovak Republic, Austria, Turkey, and Malta, out-of-field teachers seem to have higher
30 In Estonia, Hungary, and Belgium (Flemish), the difference is significant statistically.

67
 Korea
Malta
Netherlands
Poland
Turkey
Malaysia
Lithuania
Slovak Rep
Austria
Mexico
Portugal
Norway
Bulgaria
Slovenia
Denmark
Estonia
Spain
Australia
Brazil
Hungary
Belgium (F)

0 10 20 30 40 50
OFT IFT

Figure 4.10: Percentage of teachers with less-than-one-year contract by out-of-field status

number with permanent positions, but there is no detectable statistical significance in this
category. So we can conclude that out-of-field teachers tend to be hired on a short-term
basis.
One might further ask, why is this the case? My explanation is that teachers with very
short contracts probably have not established themselves within the school, or have not
achieved full certification.31 In reality, it has been commonly observed that a short contract
is used to indicate probationary status among novice teachers. If we look at the experience
level of these short-term teachers, no matter what their out-of-field status is, a majority of
them are novice teachers, with less than five years of experience.32 Therefore awarding new
teachers a short contract may be used as a screening tool while schools evaluate the new
31 This information is not currently available in TALIS.
32 Several exceptions include: Bulgaria, Denmark, Korea, Lithuania, and Mexico.

68
 Belgium (F)
Brazil
Spain
Slovenia
Hungary
Estonia
Netherlands
Australia
Portugal
Norway
Denmark
Bulgaria
Korea
Lithuania
Slovak Rep
Austria
Turkey
Mexico
Malaysia
Malta
Poland

0 20 40 60 80 100
OFT IFT

Figure 4.11: Percentage of teachers with permanent contract by out-of-field status

teachers for their competency. These teachers, who do not have a lot of job security, may
welcome any opportunity to teach that is available to them. It is not a coincidence that
out-of-field teachers in Australia, Estonia, Hungary, and Belgium (Flemish) also tend to be
young teachers. The young and inexperienced factors combined can probably explain why
these teachers who are at the bottom of school hierarchy were assigned to teach out-of-field.

4.2.5 Educational attainment

In Figure 4.12, I show the percentage of out-of-field and in-field teachers who did not
have a bachelor’s degree. Generally speaking, having a BA has become the norm in a
majority of countries regardless of out-of-field status. In 19 countries, the percentage of
in-field teachers without a BA falls below 20 percent. In 11 countries, this indicator falls
below 10 percent. Malta, Slovenia, Belgium (Flemish), and Austria are exceptions in terms

69
of their higher proportion of math and science teachers who do not have a bachelor’s degree.
Among these four countries, two of them have more in-field teachers without a BA; in the
other two countries, out-of-field teachers outnumber in-field teachers in the proportion of
non-BA degree holders.
If we look at advanced degree holders, measured by the percentage of teachers who have
a master’s degree or above (Figure 4.13), I found that the proportion of out-of-field teachers
is similar to that of in-field teachers, if not higher, who have advanced degrees. In fact, in
eight countries, there are more out-of-field teachers than in-field teachers who have advanced
degrees. They are Mexico, Belgium, Malta, the Netherlands, Norway, Spain, Poland, and
the Slovak Republic. Poland and the Slovak Republic are two extreme cases where almost
all teachers obtain a post-bachelor degree. Spain follows the same route though to a lesser
degree. This is a clear indication that out-of-field teaching is not simply a teacher supply issue
in general, as portrayed by public media. There are several explanations. One possibility is
that there could be a genuine teacher supply issue in the math and science fields because not
enough people in these fields are willing to enter teaching. The undersupply of math and
science teachers is at the same time coupled by an oversupply of non-math and non-science
graduates from teacher education programs. They are hired because of their general ability
to teach, as reflected in their advanced degree, not their competency in the subject matter.
If that is the case, one would observe a higher level of out-of-field teaching in math and
science, and a low level of out-of-field teachers in subjects such as social science, reading
and writing, and foreign language. If we recall in Figure 4.4, where the levels of out-of-field
teaching are even higher for teachers in humanities, the above explanation does not seem
logical.
A second explanation could be that the teachers with advanced degrees could have estab-
lished themselves before any kind of the subject matter competency policy is implemented;
they are recruited based on their educational qualifications and are assigned out-of-field sub-
jects. If we look at the teaching experience of these teachers in the eight countries with a

70
higher proportion of out-of-field teachers with an advanced degree in Table 4.6, it seems to
be the case for six of them. Belgium (Flemish) and the Netherlands are the exceptions, in
which advanced degree holders have little experience at all.

Australia
Denmark
Netherlands
Poland
Slovak Rep
Spain
Korea
Norway
Mexico
Portugal
Lithuania
Malaysia
Turkey
Malta
Brazil
Estonia
Bulgaria
Slovenia
Belgium (F)
Austria

0 20 40 60 80 100
OFT IFT

Figure 4.12: Percentage of teachers without a bachelor’s degree by out-of-field status

Table 4.6: Teaching experience of advanced degree holders in selected countries

Experience Malta Mexico Norway Poland Slovak Spain Belgium (F) Netherlands
<5 years 0% 0% 0% 0% 21% 15% 51% 100%
6-20 years 100% 45% 58% 100% 43% 60% 32% 0%
>20 years 0% 55% 32% 0% 35% 26% 17% 0%

This finding has implications for policies that try to alleviate out-of-field teaching by
raising qualification levels. Through the TALIS data, we see that out-of-field teachers’
attainment level does not seem to be lower than in-field teachers. A majority of out-of-field
teachers even have bachelor’s degrees or higher. To require everyone to have a master’s

71
 Brazil
Denmark
Slovenia
Malaysia
Turkey
Austria
Portugal
Australia
Lithuania
Hungary
Mexico
Belgium (F)
Malta
Netherlands
Korea
Norway
Estonia
Bulgaria
Spain
Poland
Slovak Rep

0 20 40 60 80 100
OFT IFT

Figure 4.13: Percentage of teachers with advanced degrees by out-of-field status

degree will not solve the out-of-field teaching problem, as we observed in Poland, the Slovak
Republic, and Spain, where the majority of out-of-field and in-field teachers have advanced
degrees.

4.2.6 Teaching experience

To discuss the experience level of out-of-field teachers, again I focused on the two ends of
the distribution where the differences are clear. Among inexperienced out-of-field teachers
who have less than 5 years of experience (Figure 4.15), there is a large cross-national varia-
tion. In Bulgaria, about 5 percent of its out-of-field workforce is inexperienced compared to
Belgium (Flemish), where close to 60 percent of its out-of-field teachers are inexperienced.
Three examples need some discussion: the Netherlands, Australia, and Belgium (Flemish).
In these countries, while the majority of their teaching force has ample experience, out-

72
of-field teachers are overwhelmingly inexperienced. As I have shown previously, they tend
to be young, and have less job security in school. Another five countries are also similar,
but may be to a lesser degree. Similarly, in Estonia, Brazil, Mexico, the Slovak Republic,
and Hungary, more out-of-field teachers are inexperienced than in-field teachers, though the
difference is not as large. As for the composition of veteran teachers, it is quite clear from
Figure 4.15 that there are fewer out-of-field teachers in the experienced category in these
countries.

Poland
Bulgaria
Austria
Portugal
Lithuania
Malta
Spain
Slovenia
Norway
Denmark
Korea
Hungary
Slovak Rep
Mexico
Turkey
Malaysia
Brazil
Estonia
Netherlands
Australia
Belgium (F)

0 20 40 60
OFT IFT

Figure 4.14: Percentage of teachers with less than 5 years working experience by out-of-field
status

73
 Poland
Malaysia
Australia
Malta
Brazil
Slovenia
Belgium (F)
Turkey
Portugal
Spain
Lithuania
Denmark
Slovak Rep
Korea
Estonia
Netherlands
Mexico
Hungary
Norway
Austria
Bulgaria

0 20 40 60 80
OFT IFT

Figure 4.15: Percentage of teachers with longer than 20 years working experience by out-of-
field status

4.3 Profiles of out-of-field teachers in selected countries

In the last section, I demonstrated that out-of-field teachers have different demographic
backgrounds and working conditions. There are considerable cross-country variations in the
combinations of these attributes among out-of-field teachers. In this section, I select several
countries to explore systematically the characteristics of out-of-field teachers. The benefit
of using a comprehensive dataset like TALIS is that it provides compatible statistics across
a group of very diverse countries. For this section, I exploit the cross-national feature of
TALIS, and I standardize my measures to generate comparable graphic presentations of the
attributes of out-of-field teachers in a multivariate way. I used six variables to describe the
attributes of teachers: gender, age, part-time/full-time status, highest education achieved,
years of experience, and contract length. I first computed the country mean of each indicator

74
for out-of-field and in-field teachers separately. Then I normalized each indicator across
countries. In the charts below, the unit of measure has lost its interpretable meanings due
to the ordinal or categorical nature of the data. Yet the directional feature of the data is
preserved together with the sheer difference in the magnitude of the data that makes the
findings interesting.

4.3.1 Country profile: Australia

In Australia, out-of-field teachers share similar levels of educational attainment with in-
field teachers.33 However they are relatively younger and very much inexperienced (almost
all first-year teachers), which indicates that they are at the beginning of their career. This
portrait fits with statistics collected through an Australian national survey, in which about
44 percent of beginning Australian teachers reported they had been asked to teach subjects
outside of their area(s) of qualification/expertise (Australian Education Union, 2006). It
is probably true that they were undergoing some kind of practice training stage. Australia
seems to favor flexible deployment of these young out-of-field teachers, because unlike in-
field teachers, the majority of out-of-field teachers have short contracts. Out-of-field teachers
are more likely to work on a part-time basis compared to their in-field colleagues, and it is
possible that they are undergoing a probationary period. There are also more females in
out-of-field teaching than in the regular teaching force. The gender imbalance is interesting
because it could be an indicator of a shortage of male math and science teachers in the
profession because generally they can easily find other jobs with better monetary rewards
with the same skill sets.

33 The numerical axis was not shown on Figure 4.16 due to university regulation on disser-
tation formatting. The scale of each measurement is from -2 to +2. Each line on the grid
represents a 0.5 increase.

75
 More  male  

Shorter  contract   Older  in  age  

 More  teaching  
More  part  0me  
experience  

Higher  educa0on  
IFT  
a5ainment  
OFT  

Figure 4.16: Characteristics of teachers by out-of-field status in Australia

4.3.2 Country profile: Mexico

Mexico’s case is interesting because unlike Australia, where out-of-field field teachers are
distinctive from in-field teachers, the overall attributes of out-of-field teachers in Mexico are
similar to in-field teachers.34 If we judge Mexico’s teaching force by TALIS’s standard, we
find that Mexico has the most teachers working on a part-time basis independent of their
out-of-field status. The part-time working situation seems to be more pervasive among out-
of-field teachers, though a majority of them are working on the 50 percent to 90 percent
workload of full-time teachers. Mexico’s out-of-field teachers are also much younger than
their in-field colleagues. The combination of younger age and part-time employment status
34 The numerical axis was not shown on Figure 4.17 due to university regulation on disser-
tation formatting. The scale of each measurement is from -2 to +3. Each line on the grid
represents a 0.5 increase.

76
raises a challenge to bring these teachers into the professional community in schools. The
evidence to support such a claim is that part-time out-of-field teachers are more likely to work
in multiple schools at the same time, and they do not stay very long in one particular school.
This phenomenon of moving between schools also applies to part-time in-field teachers. But
part-time out-of-field teachers have a much higher rate of teaching in multiple schools. This
suggests the transient nature of Mexico’s out-of-field teachers. They are moving from school
to school, sometimes several schools at time.

More  male  

Shorter  contract   Older  in  age  

 More  teaching  
More  part  0me  
experience  

Higher  educa0on  
IFT   a5ainment  
OFT  

Figure 4.17: Characteristics of teachers by out-of-field status in Mexico

77
4.3.3 Country profile: Belgium (Flemish)

Belgium’s out-of-field teachers are perhaps the most unique.35 They are not like in-field
teachers at all. They are the youngest and most inexperienced among the TALIS countries.
Their educational attainment is also the lowest; most of them do not have a bachelor’s degree.
Considering its in-field teaching force, who are the majority, also have lower educational
attainment, raises the question of whether Belgium does not require a college degree to be a
prerequisite for public school teachers. It is perhaps that out-of-field teachers are recruited
separately from specialized secondary teacher training institutions. Thus judging by the low
proportion of teachers who had a college degree among out-of-field and in-field teachers,
the former explanation might make more sense. However, although both out-of-field and
in-field teachers have a low attainment level, out-of-field teachers seem to be junior faculties
who are still in the probationary period because of their limited teaching experience and
shorter contract. On average, they have the shortest contract, but they work on a full-time
basis, just as in-field teachers do. There are also slightly more men in Belgium’s out-of-field
teaching force.

35 The numerical axis was not shown on Figure 4.18 due to university regulation on disser-
tation formatting. The scale of each measurement is from -2 to +3. Each line on the grid
represents a 0.5 increase.

78
 More  male  

Shorter  contract   Older  in  age  

 More  teaching  
More  part  0me  
experience  

IFT   Higher  educa0on  

a5ainment  
OFT  

Figure 4.18: Characteristics of teachers by out-of-field status in Belgium (Flemish)

4.3.4 Country profile: Norway

Norway’s out-of-field teachers are among the oldest among TALIS countries.36 Experi-
ence is the basic criterion to select out-of-field teachers. This probably explains why they
have more teaching experience than out-of-field teachers in other countries. Out-of-field
teachers are also more experienced than their in-field colleagues in Norway. Older age and
more experience suggests that these teachers do not use teaching as a second career, where
one would expect to see older teachers with less experience. Norwegian out-of-field teachers
are highly educated. Percentage-wise, more out-of-field teachers received advanced degrees
compared to in-field teachers. Their employment status, including full-time equivalence and
36 The numerical axis was not shown on Figure 4.19 due to university regulation on disser-
tation formatting. The scale of each measurement is from -2 to +2. Each line on the grid
represents a 0.5 increase.

79
contract length, are close to the out-of-field teachers in other nations. This suggests their
teaching engagement and job security are not different from in-field teachers. Norwegian
out-of-field teachers do not work at multiple schools. In fact most of them stay in one school
for a fairly long time. Thus out-of-field teachers in Norway consist of redeployed veteran
teachers from other disciplines within the same school.

More  male  

Shorter  contract   Older  in  age  

 More  teaching  
More  part  0me  
experience  

Higher  educa0on  
IFT   a5ainment  
OFT  

Figure 4.19: Characteristics of teachers by out-of-field status in Norway

4.3.5 Country profile: Turkey

Turkey’s teaching force, both in-field and out-of-field, have more males than any other
TALIS country.37 This signifies the prestige attached to teaching jobs in the public school
37 The numerical axis was not shown on Figure 4.20 due to university regulation on disser-
tation formatting. The scale of each measurement is from -2.5 to +2. Each line on the grid
represents a 0.5 increase.

80
system. Another characteristic of Turkish teachers is their age composition: they have
one of the youngest teaching forces among TALIS countries. Its in-field teaching force is
young and very inexperienced. Comparatively speaking, out-of-field teachers have slightly
more experience and are older. Since their employment status, educational attainment,
and contract type are close to in-field teachers, I suspect that the civil servant status of
the teaching job attracts more men into the profession because other job opportunities in
other sectors may not be as easily found as are government provided positions. For older
men with considerable skills and knowledge of teaching, a school job may be desirable,
even if it means to working out of one’s discipline. This could be the result of tightened
qualification standards in recent years. But such employment opportunities diminish over the
years because the male population is considerably smaller in the below 30-year-old category
among out-of-field teachers compared to in-field teachers (33 percent vs 48 percent). A
similar phenomenon is observed in the female teacher population, where there are few young
female out-of-field teachers compared to an influx of younger female in-field teachers. There
is a clear generation gap and change of composition in the teacher labor market in Turkey.

Table 4.7: Age distribution of Turkish teachers by gender and out-of-field status

Male Female
Age
OFT IFT OFT IFT
<30 years old 33% 48% 2% 50%
31-50 years old 56% 33% 98% 48%
>50 years old 11% 19% 0% 2%
Total 100% 100% 100% 100%

81
 More  male  

Shorter  contract   Older  in  age  

 More  teaching  
More  part  0me  
experience  

Higher  educa0on  
IFT   a5ainment  
OFTs  
Figure 4.20: Characteristics of teachers by out-of-field status in Turkey

4.4 Characteristics of schools where out-of-field teachers work

In this section, I compare the percentage of out-of-field teachers in mathematics and

science by various school characteristics. I exclude three countries from my analysis: Lithua-
nia, Poland, and Hungary. The reason for their exclusion is because they all have minimum
levels of out-of-field teachers in these two subject areas, all below 3 percent. The estimates
of out-of-field teaching between different types of schools are hardly detectable. To include
them in the visual presentation will not add additional information for my analysis.
Because some of the variables used in this section are constructed based on original data
provided in TALIS, I first present a definition of each variable in Table 4.8:

82
 Table 4.8: Description of school-level characteristics

Variable Definition
School location
Rural : population fewer than 3,000
Small town: population between 3,000 and about 15,000
Town: population between 15,000 and 100,000
City: population between 100,000 and about 100,000,000
Large city: population over 1,000,000
School size Small : less than 300 students
Medium: 300-1,000 students
Large: over 1,000 students
School SES Low : school percentage of parents who have ISCED 5
education are below national median AND percentage of
parents who have ISCED 3 education are below national
median
Median: 1) school percentage of parents who have ISCED
5 education are below national median AND percentage
of parents who have ISCED 3 education are above
national median; 2) school percentage of parents who
have ISCED 5 education are above national median AND
percentage of parents who have ISCED 3 education are
below national median
High: school percentage of parents who have ISCED 5
education are above national median AND percentage of
parents who have ISCED 3 education are above national
median
Classroom SES Low : classroom percentage of parents who have ISCED 5
education are below national median AND percentage of
parents who have ISCED 3 education are below national
median
Median: 1) classroom percentage of parents who have
ISCED 5 education are below national median AND
percentage of parents who have ISCED 3 education are
above national median; 2) classroom percentage of
parents who have ISCED 5 education are above national
median AND percentage of parents who have ISCED 3
education are below national median

83
 Table 4.8: cont’d

Variable Definition
Classroom SES High: classroom percentage of parents who have ISCED 5
education are above national median AND percentage of
parents who have ISCED 3 education are above national
median
Principal leadership Low : principal’s administrative leadership index is below
the 33rd percentile of national rating
Medium: principal’s administrative leadership index is
between the 33rd percentile and the 67th percentile of
national rating
High: principal’s administrative leadership index is above
the 67th percentile of national rating
Teacher shortage Low : school’s personnel shortage index is below the 33rd
percentile of national rating
Medium: school’s personnel shortage index is between the
33rd percentile and the 67th percentile of national rating
High: school’s personnel shortage index is above the 67th
percentile of national rating

4.4.1 School location

I first compare the distribution of out-of-field math and science teachers by school lo-
cations. Using the school locality information provided in the principal questionnaire, I
calculated the percentage of teachers who were teaching out-of-field for schools located in
different areas. The result is presented in Figure 4.21. It is obvious that there exist large
cross national variations corresponding to the overall between-country differences in levels
of out-of-field teaching. This is not surprising considering that TALIS includes countries at
different levels of social and economic development, which partly explains the large between-
country variations.
My interest here is to explore within-country variability. As we can see from country
to country, it is evident that schools located in rural or small town areas are prone to have
more out-of-field teachers. One of the most extreme cases is in Belgium (Flemish), where 50

84
percent38 of math and science teachers who worked in rural schools are out-of-field, compared
to less than 10 percent who worked in towns. In the rest of the countries the differences
are not so dramatic, but the pattern remains the same. This finding confirms our previous
discussion that rural schools’ ability to attract qualified math and science talents is weaker
compared to schools in more affluent suburban or metropolitan areas.

Australia Netherlands

Austria Norway

Brazil Portugal

Bulgaria Slovak Rep

Denmark Slovenia

Estonia Spain

Korea Turkey

Malaysia Belgium (F)

Malta 0 10 20 30
Mexico rural small twn
town city
0 10 20 30 large city other

Figure 4.21: Percentage of math and science teachers who are out-of-field by school location

I should remind the reader that the mean statistics used to plot the graph do not reveal
the between school variability of out-of-field teaching. Put differently, we know that out-of-
field teachers are more likely to teach in rural schools, but are they evenly distributed across
38 The figure for rural schools in Belgium (Flemish) is cropped at 30 percent in order to
maintain the display aspect ratio. Percentage of teachers who are out-of-field in math and
science in rural Belgium (Flemish) school is 50 percent. Hungary, Lithuania, and Poland
were excluded in the analyses of this section because their levels of school-level out-of-field
teaching are minimum.

85
rural schools, or are some schools more likely to deploy out-of-field teachers? In Table 4.9, I
partition the variations of school percentage of out-of-field teaching into two sources in each
country: within-school and between-schools-within-region.39 The estimates indicate that,
despite the overall higher likelihood of rural schools employing out-of-field teachers, there
are greater variations between two rural schools (between-school) in their levels of out-of-
field teachers than between a rural school and an urban school. This pattern holds across
all TALIS countries, regardless of their level of economic or educational development. It
might be easy to understand why this happens in developed and more equal nations (which
a majority of TALIS countries are), since regional development is more likely to be equalized,
which translates into school resource and teacher deployment decisions. The interesting part
is that even for developing nations such as Brazil, Malaysia, and Turkey, the between-school
variability triumphs over the between-region variability. This shows that the inequalities in
teacher staffing not only permeates across regions, but also within regions.

4.4.2 School size

The size of a school is easily connected to the type of assignment it will make. In small
schools, teachers are typically required to teach in multiple areas because such schools usually
do not have enough resources to hire subject matter specialists who only focus on a single
subject matter. The TALIS data confirms this. In Figure 4.22, I break down the percentage
of teachers who are out-of-field by the size of schools into three categories: small, medium,
and large.40 Small schools are defined to have student population of less than 300; medium
schools are defined with a student population between 300 to 1,000 students; large schools
are those with more than 1,000 students. Figure 4.22 shows that it is very common, across
39 I partition the variance by specifying a basic two-level fully unconditional model within
each country.
40 In addition to the Hungary, Lithuania, and Poland, that were excluded due to minimum
school-level out-of-field teaching, Belgium (Flemish) is also excluded in this figure because
of missing school size variable.

86
 Table 4.9: Variance partitioning of school-level OFT in math & science

Within region Between Gini 2008 GDP per

Country
between-school regions Coefficienta capita (PPP)b
Australia 93.55 6.45 30.5 38,223.58
Austria 98.33 1.67 26 39,875.63
Brazil 92.96 7.04 53.9 10,524.65
Bulgaria 97.86 2.14 45.3 13,187.39
Denmark 94.50 5.50 29 37,363.72
Estonia 94.72 5.28 30 20,319.87
Hungary 96.54 3.46 24.7 19,460.01
Korea 95.74 4.26 31 27,707.05
Lithuania 98.22 1.78 37.6 19,138.18
Malaysia 89.64 10.36 46.2 14,032.76
Malta 100.00 0.00 26 24,768.97
Mexico 94.19 5.81 51.7 14,506.46
Norway 94.04 5.96 25 52,839.78
Poland 92.87 7.13 34.2 17,592.49
Portugal 95.70 4.30 38.5 23,079.82
Slovak Rep 88.23 11.77 26 21,995.00
Slovenia 94.60 5.40 28.4 29,605.53
Spain 97.61 2.39 32 30,847.93
Turkey 87.68 12.32 40.2 13,107.55
Belgium (F) 87.18 12.82 28 36,249.42
Note: region is defined by school locations
a: CIA Worldfact book
b: measured in current dollar. Retrieved from International Monetary Fund, September 2011

countries, for small schools to have a higher percentage of their teachers to be out-of-field.
In countries such as Korea, Malaysia, and Norway, there is little difference between small
schools and medium or large schools. This suggests a further look at particular policy that
might equalize teacher distribution across schools.
Another way to look the size factor is to compare the distribution of teachers across
different school types by out-of-field and in-field teacher status. The rationale of such a
comparison is that we know small schools are more likely to have out-of-field teachers, but
due to their size, they do not necessarily deploy more out-of-field teachers in actual number.
For instance, if we assume the school-level out-of-field teaching is 10 percent for a small

87
school with 100 students and a large school with 1,000 students, then in actual number, the
small school deploys 10 out-of-field teachers and the large school deploys 100. Though 100
is greater than 10, it does not necessarily mean out-of-field situation is worse in the larger
school.
Because of the size of school is highly correlated with the number of teachers it em-
ploys, for a meaningful comparison we need to take the latter factor into consideration. In
Figure 4.23, I contrast the proportion of teachers who work in small schools by out-of-field
status. It is very clear that more out-of-field teachers, as a proportion of their total popu-
lation, work in small schools compared to in-field teachers. In four countries, more than 40
percent of out-of-field teachers are teaching in small schools: Slovenia, Estonia, Austria, and
the Slovak Republic. Here again, the concentration issue does not seem to affect Norway,
Korea, or Malaysia.

Australia Mexico

Austria Norway

Brazil Portugal

Bulgaria Slovak Rep

Denmark Slovenia

Estonia Spain

Korea Turkey

Malaysia 0 5 10 15 20
small medium
0 10 20 30 large

Figure 4.22: Percentage of teachers who are out-of-field by school size

88
 Turkey
Malaysia
Korea
Portugal
Australia
Spain
Brazil
Norway
Denmark
Mexico
Bulgaria
Slovenia
Estonia
Austria
Slovak Rep

0 20 40 60 80
OFT IFT

Figure 4.23: Proportion of teachers who work in small schools by out-of-field status

4.4.3 School socioeconomic (SES) status

In the teacher labor market literature, one concern has continued to plague researchers:
is the teacher distribution equal among students from various socioeconomic backgrounds?
Current international evidence has suggested that it is not the case in many countries (Akiba
& LeTendre, 2009; Akiba et al., 2007). Students from lower social status families are con-
sistently being assigned to teachers with less qualification and experience. For the following
presentation, I conducted a similar comparison. Since the TALIS data does not provide a
direct measure of student SES, I use parental educational background as a proxy measure
(Table 4.8). The TALIS questionnaire asked teachers to report the percentages of students’
parents who obtained ISCED 3 or ISCED 5 level of education.41 At the teacher (classroom)
41 ISCED 3 is equivalent to a high school diploma; ISCED 5 is equivalent to a bachelor’s
degree.

89
level, there were two items associated with this question. In item BTG40B, teachers were
asked to estimate the percentage of parents who were above ISCED 3-level education; in
item BTG40C, teachers were asked to estimate the percentage of parents who were above
ISCED 5-level education. I first created dummy variables using these two variables to mea-
sure whether the estimated percentage was above or below the national median. Then if the
class had parents with education levels higher than the national median at both the ISCED 3
and ISCED 5 level, I coded the classroom as “high SES classroom,” if the parental education
at the ISCED 3 and ISCED 5 level were both below national median, I coded the classroom
as “low SES classroom”; those classrooms in-between were coded “medium SES classrooms.”
I calculated a similar measure using the school-level estimates on parental education level to
reflect school SES.
I chose this coding scheme for practical reason. To illustrate my rationale, Table 4.10
shows the cross-tabulation of the two parental education dummy variables at the school
level. I found that certain proportions of schools with the amount of parents who received
ISCED 3-level education below national mean also have parents who have above-ISCED
5-level education (upper right corner). Similarly, there is also some proportion of schools
with the amount of parents who received ISCED 5-level education below national mean but
also with parents who have above-ISCED 3-level education (lower left corner). These are
probably mixed-SES schools that are difficult to be simply characterized as "low" or "high"
SES schools. Therefore I adopted a conservative strategy that only focuses on the extreme
end of the parental education distribution.
In Figure 4.24, I first show the estimated results at school level. I contrast the percentages
of teachers who were out-of-field by school SES level across countries. It is obvious that in
the majority of countries, students in low-SES schools were more likely to be taught by out-
of-field teachers in math and science. The extreme cases are Brazil, Denmark, the Slovak
Republic, and Belgium (Flemish), where out-of-field teaching in low-SES schools almost
doubled the number in high-SES schools. The highest level of out-of-field teachers in math

90
 Table 4.10: Number of schools by parental education level

Percentage of parents with Percentage of parents with above ISCED 5

above ISCED 3 education education
Below national median Above national median Total
Below national median 919 424 1343
Above national median 323 1363 1686
Total 1242 1787 3029
Note: missing data not shown

and science is found in Denmark, which was beyond 20 percent for low-SES schools. Low-
SES schools in Brazil and Estonia come the next highest, about 15 percent, in out-of-field
teaching. Interestingly for Malaysia, although the level of out-of-field teaching seemed to be
high in low-SES schools, high-SES schools did not seem to be immune from this phenomenon
either. There are visible disparities between high- and low-SES schools in the rest of the
countries, yet the overall levels are not high. Bulgaria, Malta, the Netherlands, Spain, and
Turkey actually had more out-of-field teachers in high-SES schools. These countries might
have policy that equalizes teacher distribution among high-SES and low-SES schools.
If we change our view from the school-level measure to a classroom measure, it can be
observed that in Figure 4.25, the basic pattern is preserved. Out-of-field teaching in general
was more pervasive in low-SES classrooms than in high-SES classrooms. However in the
case of Bulgaria, Spain, and Turkey, the scenario was reversed: although in general high-SES
schools have more out-of-field teachers, high-SES classes tended to have fewer out-of-field
teachers. This is due to the fact that there was more variation within schools (in terms of SES
composition) and using a school-level mean masks such variation. A classroom viewpoint
reveals that teacher sorting not only happened across schools, but also within schools.42
In the case of these three countries, the teacher’s subject matter competency is taken into
consideration when assigning a teacher (or, teachers self-select) to classes. Unanimously,
teachers with better qualification were assigned to teach well-off students. This finding
42 Studentability group might also be a factor. However, I did not have a classroom
composition measure of student SES.

91
 Australia Mexico

Netherlands
Austria
Norway
Brazil
Portugal
Bulgaria
Slovak Rep
Denmark
Slovenia

Estonia Spain

Korea Turkey

Belgium (F)
Malaysia

0 5 10 15
Malta
low SES
0 5 10 15 20 high SES

Figure 4.24: Percentage of teachers who are out-of-field by school SES

confirms the prevailing teacher sorting practice used in schools across nations.

92
 Australia Mexico

Netherlands
Austria
Norway
Brazil
Portugal
Bulgaria
Slovak Rep
Denmark
Slovenia

Estonia Spain

Korea Turkey

Belgium (F)
Malaysia

0 5 10 15
Malta
low SES
0 5 10 15 20 high SES

Figure 4.25: Percentage of teachers who are out-of-field by classroom SES

4.4.4 Principal leadership

The role that the school principal plays in making out-of-field assignments is crucial but
under-studied. Although Ingersoll (2006) suggested that unchecked discretionary power by
the school principal is a major cause of the prevalent out-of-field teaching phenomenon in
the United States, the empirical evidence on this topic is very thin.
TALIS’s principal questionnaire provided several measures of the principal’s quality, one
of which relates to the study of out-of-field teaching: a principal’s administrative leader-
ship role. In the TALIS database, this continuous variable measured two dimensions of
management components. It is a combination of accountability of management and bureau-
cratic management. Each of these aspects is measured by a set of items from the principal
questionnaire (Table 4.11). In each item, the principal rates his/her role in one of four

93
categories: “strongly disagree,” “disagree,” “agree,” and “strongly agree.” The TALIS team
first constructed a continuous index for each of the two dimensions to transform the ordinal
items into a continuous scale. Then they took the simple average of the two component
management indices and normalized composite score across countries to have an interna-
tional metric of mean zero and standard deviation one (OECD, 2010). The indices all have
high reliability and reasonable model fit. The indices “relate to school administrative tasks,
enforcing rules and procedures, and accountability role of principal” (OECD, 2010, p. 146).
For each country, I categorized schools into three levels: low-, medium-, and high-
leadership, based on the distribution of that variable. Low is defined as the 33rd percentile,
medium is the 66th percentile, and high is the highest 33rd percentile. I then calculated the
percentage of teachers who were out-of-field based on the categorization of schools.
In Figure 4.26, I present the estimated percentage of out-of-field teaching in math and
science at low- and high-leadership schools. This figure compares the level of out-of-field
teaching between the top one-third of schools in the administrative rating to the bottom
one-third of the school within each country. It is hard to argue whether there is any pattern
according to this figure. In countries such as Australia, where school management is close to
the United States’ decentralized structure, it is true that there was a higher percentage of
out-of-field teachers in low-leadership schools by a large margin. Readers should be cautioned
that the cross-country comparison based on leadership indices have two limitations. First, the
categorization of the leadership variable is made country by country in that it does not take
into account cross-country variation between schools in leadership. A “low” leadership rating
(defined as if the principal is in the lowest one-third of the leadership rating) in one country
does not necessarily mean low leadership in another country context. Secondly, because of
the way these continuous indices are constructed, having the same average leadership rating
does not necessarily reflect the same meaning across countries. It could be the case that
administrative leadership means more accountability, less bureaucracy in one country, but
not in others. The bottom line is that within-country comparison might be more meaningful.

94
 Table 4.11: Description of administrative leadership variable

Variable Indicies Items Item description

BCG16A An important part of my job is to ensure
that ministry-approved instructional
approaches are explained to new teachers,
and that more experienced teachers are
using these approaches.
BCG16D A main part of my job is to ensure that the
Accountability
teaching skills of the staff are always
role of the
improving.
principal
BCG16E An important part of my job is to ensure
that teachers are held accountable for the
attainment of the school’s goals.
BCG16F An important part of my job is to present
Administrative new ideas to the parents in a convincing
leadership way.

BCG16H It is important for the school that I see to it

that everyone sticks to the rules.
BCG16I It is important for the school that I check
for mistakes and errors in administrative
procedures and reports.
Bureaucratic BCG16J An important part of my job is to resolve
rule-following problems with the timetable and/or lesson
planning.
BCG16K An important part of my job is to create an
orderly atmosphere in the school.
BCG16O I stimulate a task-oriented atmosphere in
this school.

Interestingly, the situation was reversed in Brazil, Malaysia, and the Slovak Republic, where
high-leadership schools had a much higher level of out-of-field teachers. Mexico’s school
management system can also be characterized as centralized. As one would anticipate, there
does not seem to be too much difference in the percentage of out-of-field teaching between
these two types of schools. Does that mean that out-of-field teaching is caused by better
leadership skills?
I caution readers before making any judgment on the relationship between leadership and

95
out-of-field teaching. One limitation is that when we compare nations to nations, the sample
size becomes very small. In this case, to test the relationship between school leadership and
out-of-field teaching (and possible interaction between centralization and school leadership),
it is almost impossible to yield any generalizable findings in a statistical sense. Thus we may
characterize such relationships as illustrative at best. In addition, one might consider the
possibility of some sort of remedial policy being implemented in these countries. It is perhaps
because of the rampant level of out-of-field teachers that these high quality principals are
assigned to lead the school. Of course it is very hard to verify whether this hypothesis is
true based on a cross-sectional dataset. With more data across countries, we can see that a
principal’s influence is not the only factor that affects out-of-field teaching.

Australia Mexico

Netherlands
Austria
Norway
Brazil
Portugal
Bulgaria
Slovak Rep
Denmark
Slovenia

Estonia Spain

Korea Turkey

Belgium (F)
Malaysia

0 5 10 15
Malta
low leadership
0 5 10 15 20 high leaderhsip

Figure 4.26: Percentage of teachers who are out-of-field by principal’s administrative lead-
ership

96
4.4.5 Teacher shortage

Teacher shortage is perhaps the most commonly used argument to explain the prevalence
of out-of-field assignment. It has been argued that out-of-field assignments have been used
to fill vacancies in fields that experience teacher shortages. Does this claim have grounds? In
Figure 4.27, I present statistics based on the general personnel shortage measure in TALIS.
The variable personnel shortage in schools is available at the school level. Although it will
be ideal to have a direct measure of math and science teacher shortage, currently teacher
shortage information is not subject-matter specific.
We can tell that the out-of-field teacher level is higher in high shortage schools in Austria,
Brazil, Denmark, Mexico, the Slovak Republic, and Slovenia. Yet the difference is not large
in most of these cases between high-shortage and low-shortage schools, except for Austria,
Denmark, and the Slovak Republic. Brazil and Denmark are two alarming cases in which the
percentage of out-of-field teachers was above 20 percent in high-shortage schools. That means
one in five math and science teachers were teaching out-of-field in these countries, which is
much higher than the rest of the countries. In the rest of the countries, there were barely any
differences in the percentages of out-of-field teachers between low- or high-shortage schools.
Teacher shortages do exist within each country to varying degrees. However, it appears that
countries took different measures to alleviate the effect of teacher shortages. While out-
of-field assignments might be one strategy, we only observed it being used in a handful of
countries. Currently I lack sufficient information on whether countries have laws to regulate
out-of-field teaching in schools under a shortage. Future research might look into that issue.
Figure 4.27 explores the relationship between teacher shortages and out-of-field assignment
within countries, where the question of whether shortage is associated with high levels of
out-of-field teaching remains to be answered in the next chapter.

97
 Australia Mexico

Netherlands
Austria
Norway
Brazil
Portugal
Bulgaria
Slovak Rep
Denmark
Slovenia

Estonia Spain

Korea Turkey

Belgium (F)
Malaysia

0 5 10 15
Malta
low shortage
0 5 10 15 20 25 high shortage

Figure 4.27: Percentage of teachers who are out-of-field by school personnel shortages level

4.5 Summary

To sum up, in this chapter I present the attributes of out-of-field math and science
teachers and the schools in which they work. Cross-nationally, the overall national level
of out-of-field teaching in math and science in public schools is not very high, in all cases
below the 15-percent benchmark, except for Brazil. This finding suggests that nations are
taking a series of measures to counter the use of out-of-field teachers in math and science.
Percentages of teachers who teach out-of-field in math and science are among the lowest
compared to teachers in other subject fields. Additionally, there seems little correlation
between country-level regulation on out-of-field assignment with actual level of deployment
in math and science fields.
In terms of teacher attributes, I found that out-of-field teachers are not a homogeneous

98
group of teachers across nations. Out-of-field teachers within each nation do share similarities
with the majority of the teaching force, in-field teachers. Comparatively speaking there are
larger cross-national variations in teacher attributes, which suggests that the selection of
out-of-field teachers is highly contingent on the national context. By exploring country
by country, I discovered several modes of deployment of out-of-field teachers. Out-of-field
teachers are likely to be younger and inexperienced teachers who just start their career. In
other cases, they are older and veteran teachers with ample experience. In some cases, out-
of-field teachers are employed on a part-time basis and work at several schools at a time. The
varieties of out-of-field teachers reveal complex and context-dependent teacher assignment
strategies. Is there anything common among out-of-field teachers? I found that small, rural
and low-SES schools are most likely to deploy out-of-field teachers. This finding fits with
the previous literature, which suggests that these types of schools are least able to attract
qualified teachers, certainly including qualified math and science teachers (Ladd, 2007).
Cross-nationally, I did not find evidence that the principal’s administrative leadership might
affect out-of-field assignments. Nor does teacher shortage explain cross-national variations
in out-of-field teaching. These findings remind us that there is a long way to go before we
fully understand the mechanisms that drive out-of-field teaching. The findings presented in
this chapter only apply to a particular segment of the teacher labor force, math and science
teachers who work in public school settings. Whether or not similar observations could be
extended to teachers from other subject matter or private schools remains to be examined.

99
 Chapter 5
CROSS-NATIONAL VARIATION IN OUT-OF-FIELD TEACHING
ASSIGNMENTS AND THE DIFFERENCES BETWEEN OUT-OF-FIELD
AND IN-FIELD TEACHERS

In Chapter 4, I described some basic attributes of out-of-field teachers in TALIS countries.

The statistics shown there are mostly univariate, in the sense that each characteristic of out-
of-field teachers is used one at a time. Though convenient for graphical presentation, this
method does not take into account the impact of multiple factors at the same time. In
addition, two key questions remain unanswered:

1. What factors explain cross-national variations in out-of-field teaching?

2. Do out-of field teachers behave differently from in-field teachers in schools?

In this chapter, I answer these two questions using regression models. This chapter
is divided into two sections. In Section 5.1, I exploit cross-national variations in school
level out-of-field teachers, and I examine their relationship with three policy variables. In
Section 5.2, I compare out-of-field teachers with in-field teachers on several aspects. The
aim is to examine whether out-of-field teachers perform differently on the job and whether
or not they are given sufficient opportunity to learn.

5.1 Systemic factors explaining cross-national variation in out-of-

field teaching

In this section, I examine whether three systemic-level factors explain the cross-national
variations in out-of-field teaching.
Previous studies have primarily been in single-country settings (Ingersoll, 1999, 2003,
2004, 2006). The limitation of such an approach is obvious. Since there is little variation
within a country regarding the institutional arrangements that could potentially affect a

100
school’s decision to make out-of-field assignments, these studies were not able to examine
the relationship between country-level policy implementation and a school’s out-of-field as-
signments simply because of insufficient variability in the policy variable. Using TALIS,
which includes 21 countries, I was able to pool all these countries together in order to exploit
cross-national variations in policy settings in a wide range of countries.
I was mainly concerned with three systematic variables: the principal’s administrative
leadership, the school’s autonomy in hiring teachers, and whether or not a school is expe-
riencing teacher shortage. These three factors illustrate related aspects of teacher policy.
The principal leadership factor characterizes how schools are operated to make staffing de-
cisions. The autonomy-of-hiring variable illustrates the amount of power designated to the
school that allows them to make independent decisions in hiring staff and managing salary
schedules. The teacher shortage variable characterizes each country’s teacher labor market,
including whether enough teachers are trained to fit into available teaching positions. Ini-
tially, these factors did seem to correlate with national average levels of out-of-field teaching
to a certain degree (Figure 5.1 through Figure 5.3).

101
 .15
pct of OFT in math&sci
.1
.05
0

-2 -1 0 1
country-level mean administrative leadership
oft_sm_pub Fitted values

Figure 5.1: The relationship between country-level OFT with the principal’s administrative
leadership

102
 .15
pct of OFT in math&sci
.1
.05
0

-1 -.5 0 .5 1 1.5
school autonomy of hriing
oft_sm_pub Fitted values

Figure 5.2: The relationship between country-level OFT with the autonomy of hiring teachers

103
 .15
pct of OFT in math&sci
.1
.05
0

-1 -.5 0 .5 1
lacking personnel
oft_sm_pub Fitted values

Figure 5.3: The relationship between country-level OFT with the shortages of teachers

104
 All three of these variables were measured at the school level. Three separate continuous
measures were constructed by the TALIS team using factor analysis, and were normalized
across countries, so that these measures are comparable across countries.
The structure of this section is as follows: I first illustrate the model specifications in
section 5.1.1; in Section 5.1.2, I discuss my data, including variable descriptions and missing
data information; in section 5.1.3, I show the results and provide a discussion.

5.1.1 Empirical model specification

I specify two sets of models. The first set of models are linear models based on standard
ordinary least square (OLS) assumptions. Regression models based on Equation 5.1 through
Equation 5.6 are considered by pooling all the school observations across countries. The
outcome variable OF Tis measures the percentage of teachers who teach out-of-field in math
and science in schools i in country s. Only public schools are included in this model. The
model controls for school-level variables listed in Section 5.1.2. In one specification, I control
for country fixed effects as using dummy variables. This approach allows me to account for
all the variation in school level out-of-field teaching due to country variables. It is commonly
used in the international education literature (Ammermueller & Pischke, 2009; Hanushek
& Woessmann, 2010; Ohinata & van Ours, 2011). My primary interests are the coefficients
on three policy variables: ADM IN L, AU T HIRE, LACKP ERS, which are the principal’s
administrative leadership, the school’s autonomy in hiring teachers, and the degree to which
a school lacks instructional personnel, respectively. For each variable of interest, I fitted
two models. The first model is fitted without country dummy variables (Equation 5.1,
Equation 5.2, Equation 5.3 ) and the second model is fitted with country dummy variables
(Equation 5.4, Equation 5.5, Equation 5.6,). The idea is to compare the change estimates to
gauge how much a national context affects a school’s decision to hire out-of-field teachers. I
also repeat this strategy in a full specification, as described in Equation 5.7 and Equation 5.8,
in which I entered all variables of interest with school controls.

105
 Models without country fixed-effects:

OF Tis = δ + αADM IN Lis + γSCHOOLis + is (5.1)

OF Tis = δ + αAU T HIREis + γSCHOOLis + is (5.2)

OF Tis = δ + αLACKP ERSis + γSCHOOLis + is (5.3)

Models with country fixed-effects:

OF Tis = δ + αADM IN Lis + γSCHOOLis + as + is (5.4)

OF Tis = δ + αAU T HIREis + γSCHOOLis + as + is (5.5)

OF Tis = δ + αLACKP ERSis + γSCHOOLis + as + is (5.6)

The full model is expressed as:

OF Tis = δ + α1 ADM IN Lis + α2 AU T HIREis + α3 LACKP ERSis + γSCHOOLis + is

(5.7)

OF Tis = δ +α1 ADM IN Lis +α2 AU T HIREis +α3 LACKP ERSis +γSCHOOLis +as +is
(5.8)
Equation 5.1 through Equation 5.8 quantify a valid relationship if the assumption holds
that the relationship between variables of interest and school-level out-of-field teaching is
linear. However, because the outcome measure, percentage of out-of-field teachers in school,
is a fraction strictly between zero and one, the linear functional form in these models might

106
not be appropriate, especially at the boundaries. The fact that the outcome measure is
bounded suggests that the effect of the variable of interest cannot be constant throughout
the range of predictors. Although it is possible to include a non-linear form of predictors to
augment Equation 5.1 through Equation 5.8, there is no guarantee that the predicted values
from OLS will be lying in the unit interval (Papke & Wooldridge, 1996). As an alternative,
I fitted the same data to a set of fractional response models (FRM) introduced by Papke
and Wooldridge (1996). Specifically, I specify three non-linear models in Equation 5.9,
Equation 5.10, and Equation 5.11 without country fixed-effects:

logit(OF Tis ) = δ + αAM IN Lis + γSCHOOLis + is (5.9)

logit(OF Tis ) = δ + αAU T HIREis + γSCHOOLis + is (5.10)

logit(OF Tis ) = δ + αLACKP ERSis + γSCHOOLis + is (5.11)

I repeat the above model controlling for country fixed-effects in Equation 5.12, Equa-
tion 5.13, and Equation 5.14:

logit(OF Tis ) = δ + αAM IN Lis + γSCHOOLis + as + is (5.12)

logit(OF Tis ) = δ + αAU T HIREis + γSCHOOLis + as + is (5.13)

logit(OF Tis ) = δ + αLACKP ERSis + γSCHOOLis + as + is (5.14)

The full model is expressed as:

logit(OF Tis ) = δ+α1 ADM IN Lis +α2 AU T HIREis +α3 LACKP ERSis +γSCHOOLis +is
(5.15)

107
 and

logit(OF Tis ) = δ+α1 ADM IN Lis +α2 AU T HIREis +α3 LACKP ERSis +γSCHOOLis +as +is
(5.16)
Equation 5.9 through Equation 5.16 are non-linear models, therefore the estimated coeffi-
cients are not directly comparable to the OLS estimates. Here the quantity of interest is the
partial effect of the three independent variables on E(OF T |x). For example, the effect of ad-
ministrative leadership on out-of-field teaching can be derived by ∂E(OF T |x)/∂ADM IN L.
Papke and Wooldridge (1996) showed that Equation 5.9 through Equation 5.16 can be es-
timated using QMLE under standard generalized linear model (GLM) assumptions. They
also showed that a fully robust sandwich form estimator is computationally easy if the GLM
assumption fails.43 The QMLE estimator has been shown to have an efficiency property if
the GLM variance assumption holds. Similar to the OLS specification, I also repeat four
models (each variable of interest alone and then used together) with- or without country
dummy to the same data.
Additionally, I test the interaction term between the autonomy variable with leadership,
and autonomy with shortage variable. The models are as follows.
OLS without country fixed-effects:

OF Tis =δ + α1 ADM IN Lis + α2 AU T HIREis + α3 LACKP ERSis

+ α4 ADM IN L ∗ AU T HIRE + γSCHOOLis + is (5.17)

OF Tis =δ + α1 ADM IN Lis + α2 AU T HIREis + α3 LACKP ERSis

+ α4 LACKP ERS ∗ AU T HIRE + γSCHOOLis + is (5.18)

43 In
Stata, this estimation procedure is achieved by using glm y x1...xk, fam (bin)
link (logit) robust

108
OLS with country fixed-effects:

OF Tis =δ + α1 ADM IN Lis + α2 AU T HIREis + α3 LACKP ERSis

+ α4 ADM IN L ∗ AU T HIRE + γSCHOOLis + as + is (5.19)

OF Tis =δ + α1 ADM IN Lis + α2 AU T HIREis + α3 LACKP ERSis

+ α4 LACKP ERS ∗ AU T HIRE + γSCHOOLis + as + is (5.20)

FRM without country fixed-effects:

logit(OF Tis ) =δ + α1 ADM IN Lis + α2 AU T HIREis + α3 LACKP ERSis

+ α4 ADM IN L ∗ AU T HIRE + γSCHOOLis + is (5.21)

logit(OF Tis ) =δ + α1 ADM IN Lis + α2 AU T HIREis + α3 LACKP ERSis

+ α4 LACKP ERS ∗ AU T HIRE + γSCHOOLis + is (5.22)

FRM with country fixed-effects:

logit(OF Tis ) =δ + α1 ADM IN Lis + α2 AU T HIREis + α3 LACKP ERSis

+ α4 ADM IN L ∗ AU T HIRE + γSCHOOLis + as + is (5.23)

logit(OF Tis ) =δ + α1 ADM IN Lis + α2 AU T HIREis + α3 LACKP ERSis

+ α4 LACKP ERS ∗ AU T HIRE + γSCHOOLis + as + is (5.24)

109
 All the regression analysis used proper school weights.44

5.1.2 Data description

The three main indicators of interest are measured at the school level. As described in
Chapter 4, the administrative leadership, autonomy-of-hiring, and teacher shortage variables
were constructed from the principal questionnaire. The leadership variable was measured by
nine items that are on an ordinal scale. The autonomy of hiring variable was measured by
five items measured that are on an ordinal scale. The general personnel shortage variable
was measured by four items that are on an ordinal scale. The leadership index was developed
using confirmatory factor analysis. The autonomy of hiring and the shortage variable were
developed using principal component analysis. All of the three indices were normalized to
have an international mean of 0 and a standard deviation of 1. The description of each
variable is listed in Table 5.1.

44 InStata procedure, the data is set up with the svyset command to take into account
design replication weights. Both OLS and FRM model are estimated using svy brr pro-
cedure. The fixed-effects approach did not use svyset. Instead I applied school weights
(SCHWGT) directly in areg and glm routine.

110
 Table 5.1: Description of independent variables

Indicis Method Items Items description

Administrative Confirmatory BCG16A An important part of my job is to ensure that ministry-approved
leadership factor instructional approaches are explained to new teachers, and that more
analysisa experienced teachers are using these approaches.
BCG16D A main part of my job is to ensure that the teaching skills of the staff
are always improving.
BCG16E An important part of my job is to ensure that teachers are held
accountable for the attainment of the school’s goals.
BCG16F An important part of my job is to present new ideas to the parents in a
convincing way.
BCG16H It is important for the school that I see to it that everyone sticks to the rules.
BCG16I It is important for the school that I check for mistakes and errors in
administrative procedures and reports.
BCG16J An important part of my job is to resolve problems with the timetable
and/or lesson planning.
BCG16K An important part of my job is to create an orderly atmosphere in the
school.
BCG16O I stimulate a task-oriented atmosphere in this school.
Autonomy of Principal BCG31A autonomy in selecting teachers for hire
hiring component BCG31B autonomy in firing teachers
analysisb BCG31C autonomy in establishing teachers’ starting salaries
BCG31D autonomy in determining teachers’ salary increases
BCG31M autonomy in allocating funds for teachers’ professional development
Shortage of Principal BCG29A lack of teachers
school component BCG29B lack of laboratory technicians
personnel analysisb BCG29C lack of instructional support personnel
BCG29D lack of other support personnel
a: OECD, 2010, p. 137
b: OECD, 2010, p. 135

111
 Table 5.2: Description of missing variables

Variable Missing Total Pct missing

Administrative leadership 21 3029 0.69
Autonomy of hiring 163 3029 5.38
Shortage of school personnel 122 3029 4.03
Schoolsize 168 3029 5.55
School location 14 3029 0.46
% of parents above ISCED 5 0 3029 0.00
% of parents above ISCED 3 0 3029 0.00
principal’s experience 10 3029 0.33

The missing data information is described in Table 5.2. In general, the amount of missing
data is mild, ranging from less than 1 percent to 5.6 percent, the highest. Since missing
information is small in number, I performed list-wise deletion when fitting the models.

5.1.3 Model output

The estimates are shown in Table 5.3 through Table 5.5. First of all, the estimates be-
tween OLS and GLM (FRM) are very close, which suggests that the functional form may
not be a severe issue in estimating a school’s out-of-field assignments. Second, neither the
administrative leadership nor the teacher shortages variables show any significance in any of
the models. This is rather surprising because it suggests that across countries, the principal
leadership style and the country’s general teacher supply do not correlate with school’s deci-
sion to deploy out-of-field teachers. The only variable that allows significant interpretation
is the autonomy-of-hiring variable. The estimates suggest that if schools are given one more
standard deviation autonomy in decision-making, the percentage of out-of-field teaching will
reduce by a range of 3.1 to 3.8 percent. Considering that cross-nationally, the international
average percentage of out-of-field teaching is 9.8 percent in math and science, a 3.1 to 3.8 per-
cent reduction accounts for about 30 percent fewer out-of-field teachers. The effect is quite
substantial. In addition to the main effect of the autonomy of hiring variable, its interaction
term with the school personnel shortage variable is also significant when country fixed-effects

112
are not used (Table 5.5). The estimates for the interaction term are positive 1.7 and 2.5
percent, respectively. This shows that when schools with a greater amount-of-autonomy (1
unit increase) in making hiring decisions experience personnel shortage (1 unit increase),
they will deploy on average 1.7 to 2.5 percent more out-of-field teachers in the math and
science field. Since the autonomy and shortage variables have a standard deviation of one,
that means those schools who have 34 percent more autonomy and 34 percent more person-
nel shortage only deploy less than 3 percent more out-of-field teachers. The effects are small.
Overall, giving schools more autonomy will have net effect of 1.2 to 1.4 percentage reduction
in out-of-field teachers even when such schools experience moderate personnel shortage.
However I should point out that once the country fixed effects were accounted for, the
effects became insignificant in both the OLS and the FRM estimates. This suggests that once
the country specifics are controlled for, school autonomy of hiring teachers does not affect
a school’s percentage of out-of-field teaching, which one otherwise might expect because
the variation of autonomy might be limited within each country, therefore reducing the
association obtained cross-nationally.

113
 Table 5.3: Estimation results from school-level pooled analysis (1)

Variable OLS FRM OLS FRM OLS FRM OLS FRM

Administrative leadership 0.013 0.014 0.007 0.007
(0.010) (0.011) (0.015) (0.015)
Autonomy of hiring -0.033** -0.038* 0.010 0.011
(0.012) (0.016) (0.025) (0.024)
Lacking personnel
School sizea
Medium 0.002 -0.000 -0.039 -0.040 -0.01 -0.015 -0.045 -0.045
(0.030) (0.029) (0.036) (0.036) (0.033) (0.032) (0.038) (0.037)
Large -0.019 -0.023 -0.077+ -0.071+ -0.044 -0.046 -0.089* -0.081*
(0.030) (0.030) (0.041) (0.039) (0.034) (0.033) (0.043) (0.041)
School locationb
Small town -0.092** -0.085** -0.085* -0.076* -0.101** -0.092** -0.095** -0.085*
(0.03) (0.029) (0.034) (0.032) (0.032) (0.030) (0.036) (0.033)
Town -0.06+ -0.057 -0.057 -0.051 -0.063+ -0.056 -0.060 -0.054
(0.035) (0.035) (0.039) (0.039) (0.036) (0.036) (0.041) (0.041)
City -0.066+ -0.059 -0.055 -0.048 -0.068+ -0.061 -0.059 -0.0510
(0.035) (0.036) (0.042) (0.044) (0.037) (0.038) (0.044) (0.047)
Large city -0.107*** -0.103*** -0.104** -0.095** -0.117*** -0.110*** -0.105** -0.097**
(0.031) (0.029) (0.038) (0.034) (0.032) (0.030) (0.039) (0.035)

114
 Table 5.3: cont’d

Variable OLS FRM OLS FRM OLS FRM OLS FRM

Parental Education
20% or more above ISCED 5 0.002 -0.008 0.012 -0.010 0.010 -0.000 0.015 -0.009
(0.017) (0.022) (0.019) (0.018) (0.017) (0.023) (0.019) (0.020)
20% of more below ISCED 3 -0.050* -0.048* 0.028 0.031 -0.036+ -0.034 0.021 0.021
(0.019) (0.020) (0.024) (0.030) (0.021) (0.023) (0.024) (0.0310)
Principal’s experiencec
6-15 years -0.001 0.001 0.022 0.022 0.004 0.006 0.020 0.020
(0.025) (0.025) (0.027) (0.027) (0.026) (0.025) (0.028) (0.027)
>16 years -0.036 -0.037 0.001 0.002 -0.033 -0.034 -0.007 -0.009
(0.029) (0.031) (0.028) (0.034) (0.031) (0.034) (0.029) (0.035)
Fixed effects No No Yes Yes No No Yes Yes
N 2793 2793 2793 2793 2656 2656 2656 2656
a: Small school as referenced group
b: Rural school as referenced group
c: Principal’s experience less than 5 years as referenced group
+ p<0.10, * p<0.05, ** p<0.01, *** p<0.001

115
 Table 5.4: Estimation results from school-level pooled analysis (2)

Variable OLS FRM OLS FRM OLS FRM OLS FRM

Administrative leadership 0.012 0.011 0.006 0.006
(-0.011) (-0.012) (-0.016) (-0.016)
Autonomy of hiring -0.031* -0.036* 0.007 0.007
(-0.013) (-0.018) (-0.024) (-0.024)
Lacking personnel 0.006 0.006 -0.008 -0.009 0.000 -0.001 -0.008 -0.009
(0.011) (0.011) (0.014) (0.013) (-0.013) (-0.013) (-0.015) (-0.013)
School sizea
Medium -0.003 -0.003 -0.048 -0.050 -0.02 -0.023 -0.054 -0.056
(0.032) (0.029) (0.037) (0.038) (-0.033) (-0.032) (-0.039) (0.039)
Large -0.024 -0.027 -0.086* -0.082* -0.05 -0.051 -0.095* -0.090*
(0.031) (0.030) (0.043) (0.041) (-0.034) (-0.033) (0.045) (0.044)
School locationb
Small town -0.098** -0.091** -0.086* -0.075* -0.099** -0.090** -0.090* -0.080*
(0.031) (0.030) (0.035) (0.032) (0.033) (0.031) (0.037) (0.034)
Town -0.067+ -0.061+ -0.059 -0.050 -0.067+ -0.058 -0.063 -0.054
(0.036) (0.036) (0.040) (0.038) (0.037) (0.038) (0.042) (0.041)
City -0.067+ -0.059 -0.049 -0.037 -0.070+ -0.061 -0.0560 -0.044
(0.036) (0.037) (0.043) (0.046) (0.037) (0.037) (0.045) (0.048)
Large city -0.113*** -0.109*** -0.103** -0.093** -0.118*** -0.110*** -0.104** -0.095**
(0.032) (0.031) (0.038) (0.033) (0.034) (0.032) (0.039) (0.036)

116
 Table 5.4: cont’d

Variable OLS FRM OLS FRM OLS FRM OLS FRM

Parental Education
20% or more above ISCED 5 0.002 -0.007 0.011 -0.013 0.016 0.006 0.015 -0.011
(0.018) (0.023) (0.019) (0.019) (0.019) (0.026) (0.020) (0.021)
20% of more below ISCED 3 -0.055** -0.055** 0.024 0.026 -0.043+ -0.041+ 0.016 0.016
(0.020) (0.020) (0.025) (0.033) (0.022) (0.025) (0.026) (0.034)
Principal’s experiencec
6-15 years 0.003 0.004 0.023 0.022 0.005 0.007 0.020 0.020
(0.026) (0.025) (0.027) (0.026) (0.027) (0.026) (0.029) (0.028)
>16 years -0.037 -0.037 -0.002 -0.003 -0.032 -0.032 -0.008 -0.010
(0.030) (0.031) (0.029) (0.034) (0.033) (0.036) (0.030) (0.035)
Fixed effects No No Yes Yes No No Yes Yes
N 2699 2699 2699 2699 2559 2559 2559 2559
a: Small school as referenced group
b: Rural school as referenced group
c: Principal’s experience less than 5 years as referenced group
+ p<0.10, * p<0.05, ** p<0.01, *** p<0.001

117
 Table 5.5: Estimation results from school-level pooled analysis with interaction terms

Variable OLS FRM OLS FRM OLS FRM OLS FRM

Administrative leadership 0.012 0.014 0.008 0.009 0.012 0.011 0.006 0.006
(0.014) (0.020) (0.020) (0.026) (0.011) (0.012) (0.016) (0.016)
Autonomy of hiring -0.031* -0.037* 0.006 0.005 -0.031* -0.037* 0.005 0.005
(0.012) (0.015) (0.022) (0.022) (0.013) (0.017) (0.024) (0.024)
Lacking personnel -0.000 -0.001 -0.008 -0.009 0.012 0.021+ -0.004 -0.001
(0.013) (0.013) (0.015) (0.013) (0.010) (0.011) (0.012) (0.014)
Leadership*Autonomy -0.000 0.003 0.002 0.004
(0.013) (0.018) (0.014) (0.019)
Lacking personnel*Autonomy 0.017+ 0.025* 0.005 0.008
(0.010) (0.010) (0.010) (0.012)
School sizea
Medium -0.020 -0.023 -0.054 -0.055 -0.025 -0.030 -0.055 -0.058
(0.033) (0.032) (0.039) (0.039) (0.033) (0.032) (0.039) (0.040)
Large -0.050 -0.051 -0.095* -0.089* -0.056+ -0.059+ -0.096* -0.092*
(0.034) (0.033) (0.045) (0.043) (0.033) (0.033) (0.045) (0.044)
School locationb
Small town -0.099** -0.090** -0.091* -0.080* -0.100** -0.091** -0.090* -0.080*
(0.033) (0.032) (0.037) (0.034) (0.033) (0.031) (0.037) (0.034)
Town -0.067+ -0.059 -0.063 -0.055 -0.064+ -0.056 -0.062 -0.052
(0.037) (0.038) (0.042) (0.041) (0.037) (0.038) (0.042) (0.041)
City -0.070+ -0.061 -0.056 -0.044 -0.068+ -0.058 -0.055 -0.043
(0.037) (0.038) (0.045) (0.048) (0.037) (0.038) (0.045) (0.048)
Large city -0.118*** -0.110*** -0.104** -0.095** -0.115*** -0.108*** -0.103** -0.094**
(0.034) (0.032) (0.040) (0.036) (0.033) (0.031) (0.039) (0.035)

118
 Table 5.5: cont’d

Variable OLS FRM OLS FRM OLS FRM OLS FRM

Parental Education
20% or more above ISCED 5 0.016 0.006 0.015 -0.011 0.014 0.004 0.015 -0.011
(0.019) (0.026) (0.020) (0.021) (0.019) (0.026) (0.020) (0.021)
20% of more below ISCED 3 -0.043+ -0.041+ 0.015 0.016 -0.042+ -0.039 0.016 0.015
(0.022) (0.024) (0.026) (0.034) (0.022) (0.024) (0.026) (0.033)
Principal’s experiencec
6-15 years 0.005 0.007 0.020 0.019 0.004 0.006 0.020 0.019
(0.027) (0.026) (0.029) (0.028) (0.027) (0.026) (0.029) (0.028)
>16 years -0.032 -0.032 -0.008 -0.010 -0.028 -0.029 -0.007 -0.007
(0.033) (0.036) (0.030) (0.035) (0.033) (0.037) (0.030) (0.035)
Country fixed-effects No No Yes Yes No No Yes Yes
N 2559 2559 2559 2559 2559 2559 2559 2559
a: Small school as referenced group
b: Rural school as referenced group
c: Principal’s experience less than 5 years as referenced group
+ p<0.10, * p<0.05, ** p<0.01, *** p<0.001

119
 For the purpose of informing policymakers, it makes sense to clarify what kind of au-
tonomy can be used as policy levers. This exercise requires us to understand how the
autonomy-of-hiring variable is constructed. As described in Table 5.1, this variable is based
on five aspects of information: the responsibility of selecting teachers for hire; the responsi-
bility of firing; the responsibility of establishing teachers’ starting salaries; the responsibility
of determining teachers’ salary increases; and the responsibility of allocating funds for teach-
ers’ professional development. In each of these aspects, the principal is asked to choose from
a list of actors who is more responsible: the principal, teachers, the school governing board,
the regional or local education authority, or the national education authority. Therefore
each aspect consists of a series of zero-one binary choice. For instance, to identify who is
responsible for selecting and hiring teachers, the principal will choose from:

• The principal is more responsible (Yes/No)

• Teachers are more responsible (Yes/No)

• The school governing board is more responsible (Yes/No)

• The regional or local educational authority is more responsible (Yes/No)

• The national education authority (Yes/No)

Of course, the principal could identify multiple actors at the same time. To simplify
my analysis, I focus on each of the five actors one at time to test their relationship with
school-level out-of-field teaching. To illustrate my procedure, I ran a series of models similar
to the following:45
45 I
also replaced the administrative leadership variable and teacher shortage variables with
other items. For instance, instead of using the index of general personnel shortage, I use the
teacher shortage variable (BCG29A). The estimates are qualitatively similar, therefore I did
not present them here.

120
OF Tis = δ + α1 ADM IN Lis + α2 BCG31A1is + α3 LACKP ERSis + γSCHOOLis + as + is
(5.25)
in which the BCG31A1 variable indicates whether or not the principal is more responsible
for selecting and hiring teachers. This approach allows me to explore some of the effects of
different actors on school-level out-of-field teaching.
The findings are summarized in Table 5.6 and Table 5.7. For simplicity reasons, I show
only the estimates from the country fixed-effect specification.
Although these measures are crude in themselves, they gave me a basic sense for the
elements in school management that could relate to making out-of-field assignments. School
level out-of-field teaching is considerably higher when either teachers or regional educational
authority are given more responsibility to decide on teachers’ starting salaries. In schools
where the increase in teacher salary is determined by regional educational authority, levels of
out-of-field teaching are also higher. Additionally, when principals are given more responsi-
bility to determine the salary increase of teachers, levels of out-of-field teaching are somehow
lower. All these findings are robust to the presence of country fixed-effects.
The challenge is how to make sense of these findings. First of all, I am rather surprised to
see a significant relationship between salary schedule and school-level out-of-field teaching.
Currently there is no theory that can be readily used to explain this connection. My tentative
explanation is that school-level management is important. If the principal has more flexibility
to determine salary increase, the school might become a more desirable place to work because
it can offer more incentives to teachers. Therefore such an incentive structure could in
theory attract more high-quality teachers. This is a working hypothesis that needs more
data support. In addition, can the principal variable serve as a proxy for the unobserved
authority he/she enjoys in school? It does not seem to be the case because letting the
principal determine teachers’ starting salaries does not significantly relate statistically to
fewer out-of-field assignments.

121
 In addition, It makes little sense to jump directly to a conclusion such as “giving principals
more power to determine salary increases would lead to lower levels of out-of-field teaching.”
The items used are crude instruments that do not allow such a definitive interpretation.
The analytical approach I took also excludes the possibility that several actors might jointly
affect out-of-field teaching.46 These exploratory findings suggest that different role taking
by teachers, school principals, and local educational authority have a profound impact on
out-of-field assignments. The interaction among these three actors to determine the salary
schedule somehow leads to the decision of hiring out-of-field teachers. It implies that schools
with a school-based salary-increase mechanism are associated with lower levels out-of-field
teaching. This again confirms the relationship between out-of-field teaching and particular
school natures.

46 Without a sound theoretical framework about how these factors are interconnected with
each other, I am reluctant to try different models based on the combination of these factors.
The approach I am more inclined to, to use some indices with factor analysis, has already
been undertaken by the TALIS team and the results are shown in Table 5.3.

122
 Table 5.6: Estimation results from alternative specification (1)

Variable OLS FRM OLS FRM

Administrative leadership 0.009 0.009 0.009 0.010
(0.016) (0.015) (0.016) (0.015)
Teachers are more responsible for 0.175* 0.154***
establishing their starting salaries (0.074) (0.038)

Regional educational authority are more 0.082** 0.112**

responsible for establishing their starting (0.026) (0.035)
salaries

Lacking personnel -0.008 -0.009 -0.008 -0.009

(0.014) (0.013) (0.014) (0.013)
School sizea
Medium -0.048 -0.050 -0.048 -0.051
(0.038) (0.038) (0.037) (0.038)
Large -0.085* -0.081+ -0.084+ -0.082+
(0.043) (0.042) (0.043) (0.042)
School locationb
Small town -0.084* -0.073* -0.083* -0.071*
(0.035) (0.033) (0.035) (0.032)
Town -0.059 -0.051 -0.058 -0.048
(0.040) (0.039) (0.040) (0.038)
City -0.050 -0.038 -0.052 -0.038
(0.04) (0.05) (0.04) (0.05)
Large city -0.102** -0.093** -0.101** -0.091**
(0.039) (0.034) (0.038) (0.034)

123
 Table 5.6: cont’d

Variable OLS FRM OLS FRM

Parental Education
20% or more above ISCED 5 0.011 -0.014 0.010 -0.015
(0.019) (0.019) (0.019) (0.018)
20% of more below ISCED 3 0.022 0.024 0.019 0.021
(0.026) (0.033) (0.025) (0.032)
Principal’s experiencec
6-15 years 0.023 0.022 0.021 0.021
(0.028) (0.027) (0.028) (0.027)
>16 years -0.002 -0.003 -0.000 0.003
(0.029) (0.034) (0.028) (0.034)
Country fixed-effects Yes Yes Yes Yes
N 2651 2651 2652 2652
a: Small school as referenced group
b: Rural school as referenced group
c: Principal’s experience less than 5 years as referenced group
+ p<0.10, * p<0.05, ** p<0.01, *** p<0.001

124
 Table 5.7: Estimation results from alternative specification (2)

Variable OLS FRM OLS FRM

Administrative leadership 0.01 0.01 0.01 0.01
(0.02) (0.02) (0.02) (0.02)
Principal is more responsible for -0.043+ -0.072*
determining teachers’ salary increases (0.022) (0.036)

Regional educational authority are more 0.081** 0.102**

responsible for determining teachers’ (0.029) (0.035)
salary increases

Lacking personnel -0.008 -0.009 -0.008 -0.009

(0.014) (0.013) (0.014) (0.013)
School sizea
Medium -0.048 -0.052 -0.049 -0.051
(0.038) (0.038) (0.037) (0.038)
Large -0.087* -0.085* -0.083+ -0.079+
(0.043) (0.042) (0.043) (0.043)
School locationb
Small town -0.084* -0.072* -0.082* -0.071*
(0.035) (0.032) (0.035) (0.033)
Town -0.058 -0.046 -0.061 -0.052
(0.040) (0.039) (0.040) (0.038)
City -0.050 -0.036 -0.051 -0.037
(0.043) (0.045) (0.043) (0.046)
Large city -0.103** -0.092** -0.103** -0.093**
(0.038) (0.034) (0.038) (0.034)

125
 Table 5.7: cont’d

Variable OLS FRM OLS FRM

Parental Education
20% or more above ISCED 5 0.011 -0.013 0.016 -0.008
(0.019) (0.019) (0.019) (0.020)
20% of more below ISCED 3 0.024 0.026 0.017 0.017
(0.026) (0.033) (0.025) (0.032)
Principal’s experiencec
6-15 years 0.023 0.023 0.020 0.019
(0.028) (0.027) (0.028) (0.027)
>16 years -0.001 0.001 -0.003 -0.002
(0.028) (0.034) (0.027) (0.031)
Country fixed-effects Yes Yes Yes Yes
N 2644 2644 2645 2645
a: Small school as referenced group
b: Rural school as referenced group
c: Principal’s experience less than 5 years as referenced group
+ p<0.10, * p<0.05, ** p<0.01, *** p<0.001

126
 There are several other findings worth discussion. In Table 5.3 through Table 5.7, we can
see that several school characteristics are associated with prevalence of out-of-field teaching.
First is school size; I found that compared to small schools, large schools are less likely
to have out-of-field teaching. This effect is significant even after controlling for country-
fixed effects, which suggests that the relationship is robust across countries. The difference
between large schools and small schools varies from 7.1 percent to 9.5 percent. Second,
the rural-urban disparity again has strong implications for a school’s decision to deploy
out-of-field teachers. Compared to rural village schools, schools in large cities consistently
have a lower percentage of out-of-field teachers. The difference is substantial in all of the
cases, ranging from 9.4 percent to 11.3 percent. Third, schools with more educated parents
have fewer out-of-field teachers, although I only observed such an effect at one end of the
distribution. In schools where 20 percent or more parents have ISCED 3-above47 education,
there is a lower level of out-of-field teaching. These effects are only significant cross-nationally
without controlling for country fixed effects. They disappeared when country fixed-effects
were introduced. Again, this suggests that the distribution of out-of-field teachers across the
SES spectrum depends on the national context.
These findings suggest that the difference in levels of out-of-field teaching in math and
science are substantial across different types of schools and in schools located in different
regions. Small schools, rural schools, and low-SES schools are more likely to assign out-of-
field math and science teachers in their classrooms. This phenomenon is consistent across an
array of countries with varying degrees of educational development. The findings correspond
with those which emerged in the single country analysis I discussed in Chapter 4.
47 ISCED 3 is equivalent to high school education.

127
5.2 Comparing teaching behavior and professional status between
out-of-field and in-field teachers

In Chapter 4, I examined the attributes of out-of-field teachers, including demographic

backgrounds, working status, and school conditions. Despite the differences in these aspects
between out-of-field and in-field teachers, we still do not know whether or not they perform
differently on the job. Ideally one would like to use some productivity measures, such as
student test scores, to examine whether there is a difference in effectiveness between out-of-
field and in-field teachers in promoting student learning. However, TALIS does not include
any student test scores, and I have no way to link teachers to some sort of performance
benchmarks.48
As an alternative, I compared the teaching behaviors, classroom disciplinary climate
and professional status between out-of-field and in-field teachers. The teaching behavior
measures were used to measure whether out-of-field and in-field teacher status translates
into different instructional practices. If I find systematic differences in these aspects between
these two types of teachers, we may have reason to believe that these differences will have
differential effects on student learning outcomes. In addition, because out-of-field teachers
are believed to lack sufficient subject matter knowledge, one possible remedy would be to help
them develop essential knowledge on the job. By comparing professional status, including
access to professional development opportunities, receiving appraisals from the principal,
and collaborating with other teachers in schools, I examined whether out-of-field teachers
are treated differently in school compared to in-field teachers.
The structure of this section is as follows: I first describe the model used; I then describe
the data structure; in the last part, I show my estimation results from multiple specifications.
48 Ithas been suggested that in the next round of TALIS (2013), researchers will be able
to link TALIS with student test scores from PISA 2014.

128
5.2.1 Empirical model specification

For each country, I fitted the following OLS model:

Yis = δ + βOF Tis + πT EACHERis + as + is (5.26)

In this model, each teacher’s out-of-field status is used as an independent variable to

compare the difference between out-of-field and in-field teachers on a set of outcome measures
(represented by the generic term Yis ) as described in Section 5.2.2. Subscript i indexed
teacher and subscript s indexed school. The quantity of interest is the coefficient β associated
with the OF T variable. I controlled for a set of teacher background characteristics, including
gender, age, part-time status, contract type, education attainment, and teaching experience.
I also introduced school fixed effects as to control for the endogeneity problem. The purpose
of using school fixed effects (dummy variable) is to compare teachers within schools rather
than across diverse schools; because teachers’ time allocation and teaching practices are likely
to differ across schools. The use of fixed effects addresses any unobserved school factors that
may cause systemic difference in these practices across schools. It is commonly used in the
economics literature (Ammermueller & Pischke, 2009; Hanushek & Woessmann, 2010;
Ohinata & van Ours, 2011).
For binary outcomes, a similar logit model with school fixed-effects is specified as follows:

logit(Yis ) = δ + βOF Tis + πT EACHERis + as + is (5.27)

One concern over the fixed-effect approach is whether there are enough numbers of teach-
ers per school to provide variation among teachers. I provide a description of numbers of
teachers per school by each country in Table B.1. The majority of schools sampled in TALIS
have at least 3 to 4 faculties in math and science, which guarantees enough variation per
school. Alternatively, I ran a series of two-level hierarchical linear models (Raudenbush &
Bryk, 2001) to conduct similar exercise. For each country, the model is specified as follows:

129
 Level − 1 : Yis =δ0s + βOF Tis + π1 GEN DERis + π2 AGEis +

π3 EDLVis + π4 CON T RACTis + π5 CON T RACTis +

π6 Y REXPis + is

Level − 2 : δ0s =γ00 + γ01 LOCAT ION1s + γ02 SSIZE2s +

γ03 SSES3s + γ04 P AEXP4s + γ05 ADM IN L5s +

γ06 AU T HIRE6s + γ07 LACKP ERS7s + u0s (5.28)

where the teacher controls are exactly the same as in Equation 5.26. School controls include
location, size, SES, principal’s experience, administrative leadership, autonomy of hiring,
and whether or not school lacks personnel.49 Proper weights are used at both teacher level
and school level.50

5.2.2 Data description

As stated in Section 5.2.1, the out-of-field status variable was used as the main inde-
pendent variable. I used several outcome variables to measure various aspects of teachers’
classroom practices and their professional status. Most outcome measures are composite
indices on a continuous scale, constructed by multiple items.51 The outcome measures are
49 Itis preferable to specify a hierarchical generalized linear model (HGLM) for binary
outcomes. However, it is a challenge to incorporate multiple sampling weights in HGLM
computationally. Right now in Stata, in which I ran all the analyses for this dissertation,
only the gllamm package is capable of handling HGLM with more than one weighting variable.
In my analysis gllamm is sensitive to missing variables which results in dropping several key
variables of interest in estimation (even when the degree of missing is mild). Therefore for
operational reasons, I use only a linear probability model for binary outcomes. Also note
that all the models for Belgium (Flemish) exclude school size variable due to missing data.
50 The weights are calculated in Stata using method introduced by Pfeffermann, Skinner,
Holmes, Goldstein, and Rasbash (1998). I used the pwigls package to rescale the weights.
For technical detail, see Chantala, Blanchette, and Suchindran (2011).
51 Only two outcome measures are binary variables. I fit a similar logistic regression to
these two variables, as decribed in Equation 5.28.

130
described in Table 5.8 and Table 5.9. For a full treatment of the continuous composite in-
dices, I refer readers to TALIS 2008 Technical Report (OECD, 2010). Similar to the last
section, I include a description of missing data information in Table 5.10. The missing data
issue is mild. I performed list-wise deletion of missing cases when running the models.

131
 Table 5.8: Description of variables used as outcomes

Variable name Item Variable description

TTEACHH BTG08A Number of hours on teaching students in
school (either whole class, in groups or
individually) in a typical week
TPLANH BTG08B Number of hours planning or preparation of
lessons either in school or out of school
(including marking of student work) in a
typical week
TORDERP BTG41B Percentage of time typically spent on
keeping order in classroom (maintaining
discipline) for this target class
TTEACHP BTG41C
Percentage of time typically spent on actual
teaching and learning for this target class
CCLIMATE BTG43A, BTG43B, Classroom disciplinary climate index
BTG43C, BTG43B
TSRELAT BTG31G, BTG31H, Teacher-student relations
BTG31I, BTG31J
SELFEF BTG31B, BTG31C, Teacher self-efficacy index
BTG31D, BTG31E
TPSTRUC BTG42B, BTG43C, Classroom teaching practice index -
BTG42H, BTG42I, structuring
BTG42M
TPSTUD BTG42D, BTG42E, Classroom teaching practice index -
BTG42F, BTG42N student-oriented
TPACTIV BTG42J, BTG42O, Classroom teaching practice index -
BTG42Q, BTG42S enhanced activity
MENTOR BTG11H1 During the past 18 months, did you receive
mentoring and/or peer observation and
coaching, as part of a formal school
arrangement?
NEVERAF BTG21A, BTG21C No appraisal or feedback received by the
teacher
PDDAYS BTG12 In all, how many days of professional
development did you attend during the last
18 months?

132
 Table 5.9: Description of index variables

Variable Method of Items Item description

name construction
CCLIMATE Confirmatory BTG43A When the lesson begins, I have to wait
factor quite a long time for students to quieten
analysis down.
BTG43B Students in this class take care to create
a pleasant learning atmosphere.
BTG43C I lost quite a lot of time because of
students interrupting the lesson.
BTG43D There is much noise in this classroom
TSRELAT Confirmatory BTG31G In this school, teachers and students
factor usually get on well with each other.
analysis BTG31H Most teachers in this school believe that
students’ well-being is important.
BTG31I Most teachers in this school are
interested in what students have to say.
BTG31J If a student from this school needs extra
assistance, the school provides it.
SELFEF Confirmatory BTG31B I feel that I am making a significant
factor educational difference in the lives of my
analysis students.
BTG31C If I try really hard, I can make progress
with even the most difficult and
unmotivated students.
BTG31D I am successful with the students in my
class.
BTG31E I usually know how to get through to
students.
TPSTRUC Confirmatory BTG42B I explicitly state learning goals.
factor BTG42C I review with the students the homework
analysis they have prepared.
BTG42H I ask my students to remember every
step in a procedure.
BTG42I At the beginning of the lesson I present a
short summary of the previous lesson.
BTG42M Students evaluate and reflect upon their
own work.

133
 Table 5.9: cont’d

Variable Method of Items Item description

name construction
TPSTUD Confirmatory BTG42D Students work in small groups to come
factor up with a joint solution to a problem or
analysis task.
BTG42E I give different work to the students that
have difficulties learning and/or to those
who can advance faster.
BTG42F I ask my students to suggest or to help
plan classroom activities or topics.
BTG42N Students work in groups based upon
their abilities.
TPACTIV Confirmatory BTG42J Students work on projects that require at
factor least one week to complete.
analysis BTG42O Students make a product that will be
used by someone else.
BTG42Q I ask my students to write an essay in
which they are expected to explain their
thinking or reasoning at some length.
BTG42S Students hold a debate and argue for a
particular point of view which may not
be their own.
NEVERAF Addition BTG21A How often have you received appraisal
and/or feedback from the school
principal?
BTG21C How often have you received appraisal
and/or feedback from an external
individual or body (e.g. external
inspector)?

134
 Table 5.10: Description of missing variables

Variable name Number of Total cases Percent missing

missing data
TTEACH 558 13172 4.24
TPLANH 575 13172 4.37
TORDERP 387 13172 2.94
TTEACHP 387 13172 2.94
CCLIMATE 125 13172 0.95
TERELAT 56 13172 0.43
TPSTRUC 114 13172 0.87
TPSTUD 114 13172 0.87
TPACTIV 114 13172 0.87
TCEXCHAN 60 13172 0.46
TCCOLLAB 60 13172 0.46
SELFEF 63 13172 0.48
TBTRAD 52 13172 0.39
TBCONS 52 13172 0.39
PDDAYS 543 13172 4.12
STUAB 255 13172 1.94
MENTOR 114 13172 0.87
NEVERAF 222 13172 1.69

5.2.3 Model output

I structured this section into four parts according to different outcome variables. In
Section 5.2.3.1, I compare whether out-of-field teachers allocate different amounts of time
on various tasks compared to in-field teachers. In Section 5.2.3.2, I compare the classroom
climate and teachers’ self-efficacy between out-of and in-field teachers. In Section 5.2.3.3,
I compare whether out-of-field teachers use different teaching practices in class. In Sec-
tion 5.2.3.4, I compare the amount of professional help received by out-out-field teaching
compared to in-field teachers.
In each section, I ran six regression models using different outcome measures for each
country. The first three models use fixed-effects approach. Model specification is summa-

135
rized in Equation 5.26 and Equation 5.27.52 The next three models use HLM approach for
comparison. Model specification is summarized in Equation 5.28.53 For simplicity reasons,
the tables shown in this section only summarize the estimates of the β coefficient and its
standard error by country and by outcome measure. Estimates for control variables are
omitted.

5.2.3.1 Time on task

In Table 5.11, I first compare the time that out-of-field and in-field teachers spent on dif-
ferent tasks using the fixed-effects approach. There are no obvious patterns across countries.
The estimates are not always significant. In Denmark, Lithuania, Slovenia, and Turkey,
out-of-field teachers spend 2.3 to 7.3 hours less per week teaching students. There is little
difference in hours spent on preparing lessons. The only cases are Austria and Slovenia,
where out-of-field teachers spend 1.4 and 2 hours less per week. Percentage-wise, out-of-field
teachers in Brazil and Lithuania spend more time on keeping classroom orders; Hungary is
one exception where out-of-field teachers spend less time on keeping classroom order. For
the rest of the countries, there is essentially no statistical difference in time allocation of var-
ious tasks between out-of-field and in-field teachers. For out-of-field teachers in Denmark,
Lithuania, Slovenia, and Turkey, the general pattern is that out-of-field teachers spend less
time on learning activities and preparing for lessons, but more time on keeping classroom
order.
Table 5.12 shows estimation results from HLM models. The results are qualitatively
similar. The signs of the estimates are consistent with those of fixed-effects estimates. Out-
of-field teachers spent less time on teaching activities and preparing lessons. It is less clear
whether there is a systematic difference in keeping classroom orders across countries.

52 TheNetherlands is excluded in this analysis because the sample size is too small.
53 Malta and the Netherlands are excluded in the HLM analysis because of insufficient
sample size.

136
Table 5.11: Comparing time spent on various tasks between out-of-field and in-field teachers,
school fixed-effects approach

Number of hours Number of hours For this particular

Country on teaching planning or class, % of time on
students in school preparing of lesson keeping order
Coeff Std Error Coeff Std Error Coeff Std Error
Australia 0.449 (1.360) -1.330 (1.915) 1.920 (4.502)
Austria 0.300 (0.525) -1.401+ (0.785) 1.176 (1.764)
Brazil -1.593 (1.323) 1.591 (1.715) 3.072+ (1.690)
Bulgaria -0.223 (0.794) -0.078 (0.994) -1.506 (1.106)
Denmark -2.236+ (1.164) -1.099 (1.169) -0.224 (2.168)
Estonia -0.346 (1.306) -1.581 (1.138) -0.340 (1.728)
Hungary 0.615 (2.605) 1.166 (2.127) -4.068+ (2.261)
Korea 0.129 (1.497) 1.408 (1.367) -2.103 (2.698)
Lithuania -3.935+ (2.296) -1.516 (1.455) 8.862+ (4.924)
Malaysia 0.465 (0.649) -0.124 (0.712) 1.920 (1.567)
Malta 5.325 (3.236) 3.069 (2.785) 12.396 (8.813)
Mexico 0.718 (1.912) 0.717 (0.953) 0.397 (1.629)
Norway -0.716 (0.803) 0.506 (0.989) -1.189 (1.794)
Poland 0.129 (9.109) 1.921 (3.244) 3.558 (3.120)
Portugal 0.381 (2.120) 0.300 (1.406) -1.412 (4.079)
Slovak Rep -1.278 (1.406) 0.565 (1.393) -0.431 (2.451)
Slovenia -2.452+ (1.356) -2.071* (0.989) 1.357 (1.804)
Spain -0.566 (0.525) -0.997 (0.810) 1.106 (2.284)
Turkey -7.349** (2.689) -3.435 (2.679) 5.796** (2.002)
Belgium (F) 0.675 (0.802) 2.277 (1.873) 5.224 (3.843)
Note: Controlled for gender, age, part-time status, contract type,
education attainment, and teaching experience

137
Table 5.12: Comparing time spent on various tasks between out-of-field and in-field teachers,
HLM approach

Number of hours Number of hours For this particular

Country on teaching planning or class, % of time on
students in school preparing of lesson keeping order
Coeff Std Error Coeff Std Error Coeff Std Error
Australia 0.592 (0.958) -0.274 (1.849) 3.084 (4.186)
Austria 0.570 (0.546) -1.503+ (0.863) -0.273 (1.539)
Brazil -1.368 (1.345) 2.182+ (1.154) 0.806 (1.538)
Bulgaria 0.002 (1.117) -0.046 (0.948) 0.334 (0.839)
Denmark -1.845* (0.836) -0.838 (0.926) 4.213+ (2.435)
Estonia -0.304 (1.105) -2.479** (0.927) -0.160 (1.226)
Hungary -0.360 (1.477) -0.520 (1.338) -2.167 (2.361)
Korea 0.612 (1.455) 2.382* (1.088) -2.422 (2.068)
Lithuania -3.776+ (2.081) -2.208+ (1.202) 8.141 (6.056)
Malaysia 0.732 (0.653) 0.065 (0.712) 2.071 (1.497)
Mexico -1.227 (2.152) 0.590 (1.017) -0.570 (1.656)
Norway -0.379 (0.717) 0.453 (0.821) -1.256 (1.545)
Poland -5.084*** (1.151) -4.223*** (0.954) -6.052*** (1.163)
Portugal 1.533 (1.618) -1.338 (1.434) 1.016 (3.967)
Slovak Rep -1.548 (1.050) -1.017 (1.257) -1.004 (1.763)
Slovenia -2.207+ (1.155) -2.524** (0.914) 1.589 (1.924)
Spain -0.603 (0.634) -1.277 (0.827) -2.600 (1.859)
Turkey -3.168* (1.608) 0.532 (1.969) 3.025 (2.663)
Belgium (F) 0.100 (0.655) 1.392 (1.158) 3.172 (3.826)
Note: Controlled for gender, age, part-time status, contract type,
education attainment, teaching experience, school location, sizes, SES,
administrative leadership, personnel shortage, autonomy of hiring, and
principal’s years of experience.

138
5.2.3.2 Disciplinary climate

One explanation of the differences in time spent is that out-of-field teachers are assigned
to tough classrooms and their limited ability to control in classrooms. There seems to be
some grounds for this hypothesis. The estimates are listed in Table 5.13 and Table 5.14. In
Austria, Brazil, Lithuania, Malaysia, Poland, and Turkey, the classroom disciplinary climate
is consistently worse for out-of-field teachers. The outcome measure, classroom disciplinary
climate, is standardized to have a zero mean and a standard deviation of one. We can thus
interpret the outcome in terms of standard deviation. This finding suggests that out-of-field
teachers’ classroom climate is 0.24 to 0.55 standard deviation lower than in-field teachers’.
The difference in classroom climate also translates into the teacher-student relationship. In
Brazil, Denmark, Mexico, Poland, Portugal, the Slovak Republic, Spain, and Turkey, out-of-
field teachers are likely to have worse relationships with their students compared to in-field
teachers. The difference is still in the range of 0.18 to 0.55 standard deviation. The estimates
from HLM analysis are relatively smaller and insignificant.
It is less clear about what causes such a difference in classroom climate between out-
of-field and in-field teachers. It could be that out-of-field teachers have worse classroom
management skills so that they cannot control the classroom and students. On the other
hand, although school fixed-effects guarantee that cross-school sorting is taken into account,
it is still possible that out-of-field teachers are assigned to teach in tougher classrooms. As
seen in Chapter 4, I found some partial evidence for this claim because we have seen that
out-of-field teachers are more likely to be assigned to low-SES classrooms in each school and
students’ SES status could affect teachers’ perception of classroom discipline.
Although some out-of-field teachers teach in less-disciplined classrooms, this does not
seem to affect the way out-of-field teachers feel about themselves. They feel the same about
their role in changing student learning outcomes in most cases. There are only three cases,
Mexico, Spain, and Belgium (Flemish), where out-of-field teachers felt much worse about
their job situation, in the range of 0.32 to 0.69 standard deviation, compared to in-field

139
teachers.

Table 5.13: Comparing classroom management between out-of-field and in-field teachers,
school fixed-effects approach

Classroom
Teacher-student Teacher’s self
Country disciplinary
relationship index efficacy index
climate index
Coeff Std Error Coeff Std Error Coeff Std Error
Australia 0.123 (0.302) 0.030 (0.196) 0.211 (0.358)
Austria -0.261+ (0.148) -0.015 (0.147) 0.088 (0.130)
Brazil -0.277* (0.120) -0.231+ (0.132) -0.175 (0.157)
Bulgaria 0.071 (0.118) -0.260 (0.181) 0.166 (0.186)
Denmark 0.191 (0.191) -0.442* (0.211) 0.040 (0.184)
Estonia -0.146 (0.186) -0.054 (0.151) 0.046 (0.093)
Hungary 0.160 (0.374) 0.255 (0.341) 0.689* (0.323)
Korea 0.248 (0.217) -0.021 (0.169) 0.126 (0.138)
Lithuania -0.555* (0.220) -0.174 (0.226) -0.354 (0.255)
Malaysia -0.227+ (0.118) 0.016 (0.092) 0.009 (0.110)
Malta -0.542 (0.460) -0.052 (0.431) -0.155 (0.382)
Mexico 0.118 (0.128) -0.292+ (0.170) -0.439** (0.165)
Norway 0.172 (0.230) 0.077 (0.144) 0.034 (0.165)
Poland -0.531* (0.226) -0.176+ (0.096) 0.067 (0.101)
Portugal -0.102 (0.242) -0.285+ (0.154) -0.294 (0.188)
Slovak Rep 0.174 (0.285) -0.470** (0.154) -0.322 (0.277)
Slovenia 0.147 (0.284) -0.013 (0.107) -0.160 (0.134)
Spain -0.050 (0.190) -0.257* (0.124) -0.322+ (0.173)
Turkey -0.526* (0.215) -0.546+ (0.302) -0.409 (0.412)
Belgium (F) -0.075 (0.326) 0.107 (0.208) -0.687+ (0.372)
Note: Controlled for gender, age, part-time status, contract type,
education attainment, and teaching experience

140
Table 5.14: Comparing classroom management between out-of-field and in-field teachers,
HLM approach

Classroom Teacher-student Teacher’s self

Country disciplinary relationship index efficacy index
climate index
Coeff Std Error Coeff Std Error Coeff Std Error
Australia -0.044 (0.252) 0.038 (0.229) -0.186 (0.220)
Austria -0.162 (0.166) -0.168 (0.132) 0.027 (0.134)
Brazil -0.130 (0.089) -0.135 (0.090) -0.138 (0.121)
Bulgaria -0.016 (0.108) -0.136 (0.120) 0.225 (0.168)
Denmark -0.015 (0.154) -0.061 (0.169) 0.032 (0.176)
Estonia -0.134 (0.158) -0.013 (0.128) -0.056 (0.089)
Hungary -0.210 (0.392) 0.347 (0.364) 0.471 (0.304)
Korea 0.038 (0.170) -0.121 (0.122) 0.122 (0.104)
Lithuania -0.287 (0.262) -0.249 (0.210) -0.241 (0.252)
Malaysia -0.268* (0.116) -0.051 (0.094) 0.029 (0.110)
Mexico -0.003 (0.119) -0.269 (0.174) -0.466** (0.165)
Norway 0.115 (0.223) -0.016 (0.122) -0.103 (0.123)
Poland -0.038 (0.137) -0.231+ (0.130) -0.131 (0.118)
Portugal -0.237 (0.181) -0.271 (0.167) -0.157 (0.219)
Slovak Rep 0.421+ (0.237) -0.222* (0.105) -0.178 (0.178)
Slovenia -0.014 (0.234) 0.021 (0.090) -0.113 (0.110)
Spain 0.179 (0.173) -0.182 (0.122) -0.093 (0.176)
Turkey -0.423* (0.214) -0.382+ (0.212) -0.283 (0.243)
Belgium (F) 0.252 (0.238) -0.152 (0.204) -0.705* (0.301)
Note: Controlled for gender, age, part-time status, contract type,
education attainment, teaching experience, school location, sizes, SES,
administrative leadership, personnel shortage, autonomy of hiring, and
principal’s years of experience.

141
5.2.3.3 Teaching practice

Ingersoll (1999) suggested that out-of-field teachers are more likely to use structured
and routine practices because they do not have sufficient pedagogical content knowledge to
tailor their teaching to students with different needs. Is this true for out-of-field teachers in
TALIS countries? The estimates are listed in Table 5.15 and Table 5.16. From the results
of the fixed-effects approach, it does not appear to be the case. In fact, out-of-field teachers
almost unanimously used less-structured practice in classroom (though the difference not
always significant). There is only one case, Hungary, where out-of-field teachers did use
more structured lecturing. Similarly, out-of-field teachers used less student-oriented teaching
practices and enhanced activities. There are exceptions, though. In Malta and Slovenia, out-
of-field teachers reported using more student-oriented practices and enhanced activities in
the classroom.
Estimates from the multi-level approach generally shows no difference in structuring prac-
tices between out-of-field and in-field teachers except for two countries. Out-of-field teachers
in Estonia use less structured teaching practices. Out-of-field teachers in Poland use more
structured teaching practice. There is hardly any differences in student-oriented practices
between out-of-field and in-field teachers. In terms of using enhanced activities, though out-
of-field teachers in 12 countries appear to use less structured teaching practices, only two
are artistically significant. Surprisingly, out-of-field teachers in Australia and Denmark use
more enhanced activities, which are not consistent with the results from the fixed-effects
approach.

142
Table 5.15: Comparing teaching practices between out-of-field and in-field teachers, school
fixed-effects approach

Teaching practice: Teaching practice: Teaching practice:

Country Structuring Student-oriented Enhanced
activities
Coeff Std Error Coeff Std Error Coeff Std Error
Australia -0.125 (0.185) -0.269 (0.190) -0.290* (0.137)
Austria 0.013 (0.132) 0.154 (0.105) 0.069 (0.062)
Brazil -0.048 (0.169) -0.254+ (0.148) -0.239 (0.150)
Bulgaria -0.236* (0.109) -0.204 (0.165) -0.117 (0.128)
Denmark -0.247* (0.114) -0.299** (0.114) -0.259* (0.109)
Estonia 0.044 (0.116) 0.061 (0.112) -0.026 (0.079)
Hungary 0.459+ (0.261) 0.088 (0.232) 0.175 (0.177)
Korea -0.248 (0.261) -0.099 (0.218) -0.029 (0.256)
Lithuania -0.288 (0.337) 0.135 (0.244) 0.190 (0.174)
Malaysia 0.011 (0.125) 0.099 (0.114) 0.067 (0.115)
Malta 0.441 (0.274) 0.726+ (0.369) 0.300+ (0.176)
Mexico -0.300+ (0.173) -0.417** (0.132) -0.492*** (0.148)
Norway -0.125 (0.111) 0.047 (0.077) 0.056 (0.077)
Poland 0.537 (0.497) 0.507*** (0.104) -0.025 (0.411)
Portugal -0.247 (0.243) -0.149 (0.214) -0.208 (0.148)
Slovak Rep -0.137 (0.217) -0.112 (0.208) -0.042 (0.117)
Slovenia 0.108 (0.177) 0.054 (0.122) 0.101+ (0.053)
Spain -0.011 (0.130) 0.046 (0.094) 0.032 (0.087)
Turkey -0.440 (0.351) -0.485* (0.232) -0.476+ (0.267)
Belgium (F) -0.368 (0.239) 0.063 (0.261) 0.021 (0.228)
Note: Controlled for gender, age, part-time status, contract type,
education attainment, and teaching experience

143
Table 5.16: Comparing teaching practices between out-of-field and in-field teachers, HLM
approach

Teaching practice: Teaching practice: Teaching practice:

Country Structuring Student-oriented Enhanced
Coeff Std Error Coeff Std Error activities
Coeff Std Error
Australia -0.338* (0.153) -0.495*** (0.116) -0.405*** (0.086)
Austria -0.084 (0.121) 0.175+ (0.092) 0.058 (0.048)
Brazil -0.105 (0.112) -0.075 (0.119) -0.060 (0.122)
Bulgaria -0.018 (0.095) -0.057 (0.135) 0.022 (0.105)
Denmark -0.313** (0.104) -0.202+ (0.110) -0.146 (0.109)
Estonia 0.048 (0.086) 0.040 (0.090) -0.021 (0.063)
Hungary 0.211 (0.229) 0.183 (0.173) 0.211 (0.187)
Korea -0.021 (0.224) -0.079 (0.149) -0.021 (0.177)
Lithuania -0.282 (0.352) -0.032 (0.291) 0.113 (0.201)
Malaysia 0.030 (0.135) 0.144 (0.105) 0.087 (0.100)
Mexico -0.254 (0.171) -0.470** (0.160) -0.568*** (0.166)
Norway -0.156 (0.095) -0.002 (0.069) 0.003 (0.065)
Poland -0.368* (0.152) 0.224* (0.111) -0.079 (0.125)
Portugal -0.189 (0.205) -0.093 (0.195) -0.129 (0.123)
Slovak Rep -0.047 (0.218) 0.103 (0.191) 0.029 (0.112)
Slovenia -0.147 (0.136) -0.063 (0.118) 0.014 (0.056)
Spain 0.015 (0.129) 0.074 (0.105) 0.059 (0.100)
Turkey 0.009 (0.193) -0.112 (0.187) -0.089 (0.194)
Belgium (F) -0.360 (0.221) 0.031 (0.264) 0.059 (0.211)
Note: Controlled for gender, age, part-time status, contract type,
education attainment, teaching experience, school location, sizes, SES,
administrative leadership, personnel shortage, autonomy of hiring, and
principal’s years of experience.

144
5.2.3.4 Professionl status

Regarding out-of-field teacher’s professional status, the estimates are listed in Table 5.17
and Table 5.18. Under the school fixed-effects approach, I use logistic regression to fit
the mentoring and appraisal models. The professional development model is fitted using
OLS. Therefore the point estimates for the second and fourth columns in Table 5.17 are
exponentiated coefficients for logistic regression models. In the HLM approach, I only use
linear probability models to fit the mentoring and appraisal models. Therefore the estimates
in Table 5.18 should be interpreted in probability terms.
In Estonia and Slovenia, they appear to receive a little bit more mentoring and peer
observation than in-field teachers from the fixed-effects estimates. However the effects are not
consistent with the HLM estimates. In the latter case, out-of-field teachers in Estonia receive
less mentoring and peer observation compared to in-field teachers. There is no statistical
difference in the amount of mentoring and peer observation between Slovenian teachers from
HLM estimates.
The estimates from the fixed-effects approach show that in Australia, Brazil, and Bul-
garia, out-of-field teachers are less likely to receive feedback about their work from their
principal or external inspector. For Australian out-of-field teachers, the situation is espe-
cially worse. They are nearly 10 times less likely to receive any feedback compared to their
in-field peers. In Lithuania, out-of-field teachers spend less time (4.8 days) on professional
development. Again, the HLM estimates are smaller and the significant estimates did not
always line up with those from the fixed-effects approach. Yet even after taken these dif-
ference into consideration, it is easy to find that the statistically significant estimates are
special cases. All the other non-significant findings paint a picture that out-of-field teachers
are not treated differently in school compared to their in-field peers.
Is it a good thing or bad thing? It is hard to say. On the one hand, it might be good
that these out-of-field teachers are not being singled out from their peers. Yet one might
argue that they in fact need more help to develop because they lacked certain knowledge

145
when they first started. If there is no remedial policy targeting them, how can we expect
that they will just learn by themselves?
Though not all the findings are statistically significant across TALIS countries, they
illuminate a not-so-rosy picture of out-of-field teachers in certain countries. For instance,
the Brazilian out-of-field teachers spend about 3 percent more time on maintaining classroom
order. Their classroom climate is about 0.3 standard deviation worse than in-field teachers.
They are also less likely to use student-oriented practices and yet they are twice as likely to
not receive any appraisals. Out-of-field teachers in Lithuania also spend considerably less
time on teaching (4 hours) and more on keeping classroom order (8.9%). Their classroom
climate is half-a-standard-deviation worse than in-field teachers’. They spend 4.8 days less
on professional activities than in-field teachers do. Turkish out-of-field teachers spend the 7
hours less on teaching and 5.8 percent more on keeping order. Yet their classroom climate is
half-a-standard-deviation worse than in-field teachers, and that teacher-student relationships
are similar. They are less inclined to use a student-oriented approach or enhanced activity
in teaching (0.49 and 0.48 standard deviation, respectively).

146
Table 5.17: Comparing professional status between out-of-field and in-field teachers, school
fixed-effects approach

Receiving Never received

Days spent on
mentoring or peer appraisal or
Country professional
observation feedback
development
(probability) (probability)

Coeff Std Error Coeff Std Error Coeff Std Error

Australia 0.040 (0.150) 0.075 (0.077) 0.963 (1.223)
Austria -0.029 (0.052) 0.024 (0.051) -0.922 (0.919)
Brazil -0.049 (0.060) 0.056 (0.054) -0.629 (3.548)
Bulgaria 0.066 (0.075) 0.017 (0.026) -33.105 (30.284)
Denmark -0.024 (0.067) 0.133+ (0.070) 3.059 (3.338)
Estonia -0.193** (0.064) -0.048+ (0.025) 0.291 (2.155)
Hungary 0.146 (0.173) 0.011 (0.122) 0.769 (5.113)
Korea -0.079 (0.109) 0.059 (0.072) 1.127 (5.963)
Lithuania -0.117 (0.144) 0.044 (0.055) -4.831+ (2.855)
Malaysia -0.005 (0.055) -0.020 (0.015) -0.610 (1.706)
Malta 0.331* (0.133) 0.501** (0.172) 8.375 (7.015)
Mexico -0.135 (0.083) -0.044 (0.031) 1.080 (15.580)
Norway -0.081 (0.067) -0.015 (0.051) -0.532 (2.404)
Poland -0.031 (0.063) -0.418 (0.340) -7.573 (7.727)
Portugal -0.024 (0.107) -0.003 (0.133) -26.589 (16.390)
Slovak Rep -0.083 (0.139) 0.044 (0.047) -1.258 (1.846)
Slovenia -0.147+ (0.078) 0.053 (0.050) 0.483 (1.909)
Spain -0.025 (0.058) -0.047 (0.084) -5.755 (3.768)
Turkey -0.142 (0.202) -0.085 (0.072) -5.650 (3.664)
Belgium (F) -0.136 (0.116) 0.073 (0.088) 1.395 (5.566)
Note: Controlled for gender, age, part-time status, contract type,
education attainment, and teaching experience.

147
Table 5.18: Comparing professional status between out-of-field and in-field teachers, HLM
approach

Receiving Never received

Days spent on
mentoring or peer appraisal or
Country professional
observation feedback
development
(probability) (probability)

Coeff Std Error Coeff Std Error Coeff Std Error

Australia -0.149 (0.099) 0.070 (0.088) 2.180+ (1.278)
Austria 0.008 (0.065) -0.053 (0.041) -1.297 (1.041)
Brazil -0.003 (0.049) 0.036 (0.042) -0.782 (2.727)
Bulgaria 0.102 (0.077) 0.053 (0.046) 1.110 (9.075)
Denmark -0.043 (0.057) -0.009 (0.039) 7.592* (3.022)
Estonia -0.109* (0.054) -0.020 (0.021) -0.581 (1.666)
Hungary -0.101 (0.120) 0.021 (0.060) 2.257 (3.692)
Korea -0.026 (0.107) 0.063 (0.067) 4.523 (5.622)
Lithuania 0.005 (0.151) 0.096 (0.081) -1.918 (1.695)
Malaysia -0.036 (0.043) -0.016 (0.013) 1.340 (1.468)
Mexico -0.065 (0.077) -0.057** (0.021) -0.330 (12.814)
Norway -0.059 (0.057) -0.031 (0.048) -1.013 (1.103)
Poland 0.209** (0.070) -0.012 (0.040) -13.668+ (7.203)
Portugal -0.034 (0.081) -0.035 (0.119) -14.185 (9.451)
Slovak Rep -0.098 (0.112) 0.019 (0.035) -1.033 (1.328)
Slovenia -0.065 (0.084) 0.054 (0.055) -0.347 (1.785)
Spain 0.002 (0.064) -0.120 (0.091) -7.024* (3.049)
Turkey -0.121 (0.087) 0.039 (0.072) 1.591 (2.802)
Belgium(F) -0.125 (0.081) 0.052 (0.060) -1.98 (3.838)
Note: Controlled for gender, age, part-time status, contract type,
education attainment, teaching experience, school location, sizes, SES,
administrative leadership, personnel shortage, autonomy of hiring, and
principal’s years of experience.

148
5.3 Summary

For the results presented in this chapter, I used regression methods to study two ques-
tions: 1) What factors are associated with cross-national variations in school-level out-of-field
teaching? 2) On what aspects do out-of-field teachers differ from in-field teachers?
Regarding the first question, I find that school autonomy in hiring teachers is correlated
with school-level out-of-field teaching in math and science. Schools with higher autonomy
generally have lower level out-of-field teaching in math and science. The relationship is
highly contingent upon national context. Once the country fixed-effects are controlled for,
the significance disappear. Additionally, other school characteristics such as location, size,
and SES, also significantly correlated with levels of out-of-field teaching. Rural schools and
small schools consistently deploy more out-of-field teachers. Low-SES schools also tend to
have more out-of-field teachers, but the effects are only significant cross-nationally. These
findings are consistent with previous literature on the teacher labor market which concludes
that these types of schools have a disadvantage in hiring high-quality teachers.
The picture of the performance of out-of-field teachers is mixed. In a small number of
countries, out-of-field teachers spend less time in teaching and preparing lessons, and more
time in keeping classroom order. For the majority of countries, the difference is not significant
after accounting for teacher attributes and school factors. There is also some evidence that
out-of-field teachers are assigned to classrooms with worse climates and out-of-field teacher’s
self-reported relationship with students therefore are worse. However out-of-field teachers
do not have lower self-efficacy. Contrary to previous studies, out-of-field teachers used less
structured teaching practices in the classroom. They are also slightly less likely to use
student-oriented practices and enhanced activities in teaching.
Lastly, out-of-field math and science teachers are not given extra attention on the job.
They do not receive more opportunities for teacher collaboration, and they do not receive
more feedback on their work. Their time spent on professional development activity is also
close to that of in-field teachers. In the next chapter I will discuss the implications of these

149
findings for policy and administration.
However these findings should be interpreted with caution. Regression estimates from
two different approaches do not always yield the same conclusion. The fixed-effects approach
could suffer from insufficient variation within schools thus leading to inflated standard er-
rors. The HLM multilevel approach, although not limited by number of teachers per school,
suffers from potential omitted variable bias. School-level information such as curriculum
arrangement, length of school day, and resources devoted to professional development is not
directly observed and therefore is not explicitly controlled in HLM models. These factors
could potentially cause bias in HLM estimates.

150
 Chapter 6
DISCUSSION AND CONCLUSION

6.1 Summary of findings

The primary goal of this study is to provide a basic understanding of the nature of
out-of-field assignments in today’s classrooms. Unlike previous studies which built upon
nationally representative teacher data, this is the first study that utilize such information
from a comparative perspective. Though researchers and practitioners have long realized
that out-of-field teaching is commonly used in the classroom, this is the first empirical study
that provides comprehensive statistics on this issue. This study focuses on math and science
teachers who work in public schools. I tried multiple ways to provide a detailed portrait of
mathematics and science teachers who are not trained in the subjects they are training.
The findings are multifaceted. First of all, for most TALIS countries, out-of-field teaching
is a commonly used strategy in classrooms in math and science subjects. The overall degree
of out-of-field teaching is moderate, compared to what has already been found in the United
States. On average, about one in ten teachers in public math and science classrooms is
teaching out-of-field. There are large between-country variations in out-of-field teaching. In
countries such as Hungary, Lithuania, and Poland, the out-of-field teacher is extremely rare,
consisting of fewer than 3 percent of the entire teaching force; on the other hand, countries
such as Brazil, Estonia, and Turkey deploy larger numbers of out-of-field teachers, with over
10 percent of their teaching force being out-of-field. I should note that even the country
with the most severe out-of-field teaching in TALIS fell well below the estimates for the
United States. Moderate levels of out-of-field teaching among TALIS countries show that
out-of-field teaching has been effectively limited in many countries around the world. The
commitment to building a high quality math and science teaching force by ensuring every
teacher has sufficient qualification is reflected in these statistics.

151
 So who are these out-of-field teachers? I spent considerable effort trying to investigate
the attributes of out-of-field teachers, hoping to find certain patterns that could characterize
the whole out-of-field teaching force. Yet my analysis shows that out-of-field teachers are
not a homogenous group of teachers. It makes little sense to conclude with one image of out-
of-field teachers. For one thing, I found that out-of-field teachers are much like their in-field
colleagues in each country. There is more similarity between out-of-field and in-field teachers
within each country than out-of-field teachers shared with each other across borders. This
suggests that just like the attributes of the teacher workforce around the world that differ
from country to country, out-of-field teachers are largely determined by national policies and
regulations. This finding corresponds to the earlier statement that national context plays an
important role in shaping the teacher labor force.
Though the characteristics of out-of-field teachers vary across nations, there are certain
patterns of similarities among out-of-field teachers; notably the combination of demographic
and occupational factors. For example, younger out-of-field teachers are also likely to be
inexperienced teachers with less job security. Older out-of-field teachers also have ample
teaching experience and a lot of job security. There are more countries deploying young and
inexperienced out-of-field teachers than older and experienced out-of-field teachers. Across
countries, out-of-field teachers have equal amounts of education compared to their in-field
colleagues. They are more likely to work on a part-time basis.
In terms of types of schools in which they work, math and science out-of-field teachers
across countries find strikingly more similarities in their working place. I found that out-of-
field teachers of these two subjects are more likely to work in rural schools, small schools,
and low-SES schools. This pattern is consistent across countries and also holds within most
countries. The concentration of out-of-field teachers in these types of schools reflects the
unequal distribution of teachers in several countries. This finding reveals the complex nature
of teacher distribution across schools. The national average level of out-of-field teaching is
likely to mask the within-country between-school variations. For instance, in addition to the

152
high overall level of out-of-field teaching in Denmark (about 15 percent nationally), small
schools have almost 30 percent out-of-field teachers. Variations between schools should also
not be overlooked. There are also rare cases where the teacher distribution is more equal
between schools. For example in Bulgaria, Korea, Lithuania, Mexico, Norway, Portugal
and Spain, the regional differences in levels of out-of-field teaching are much smaller. Across
schools of different sizes, the levels of out-of-field teaching are more equal in Bulgaria, Korea,
and Turkey. At the classroom level, Bulgaria, Korea, Mexico, and Slovenia also ensure that
out-of-field teacher distribution is equal among high and low SES classrooms.
I also found that an increase a school’s autonomy in hiring and staffing teachers could
reduce out-of-field teaching in math and science across countries. Two other factors, a
school’s administrative leadership and teacher shortage experienced by the school do not
seem to explain the variation in out-of-field teaching.
Among all the TALIS countries, Poland has the lowest level of out-of-field teaching in
math and science, and Brazil and Denmark have the highest level. Distribution-wise, Bul-
garia, Korea, and Mexico have a more equal distribution of out-of-field teachers among
schools with different characteristics. Levels of out-of-field teaching in math and science are
also low in these three countries. In Australia, Brazil, Denmark, and Estonia, not only are
the levels of out-of-field teaching in math and science higher, but the distribution is highly
unequal. Disadvantaged schools consistently deploy more out-of-field teachers.

6.2 Implication of policy and administration

There are several policy implications that can be derived from this study. First of all, out-
of-field teaching is not just a general undersupply issue.54 I found that out-of-field teaching
54 There is a nuanced difference between the conventional use of shortages and shortages
in the broad sense in economics. Any imbalance between labor demand and supply can be
labeled as a shortage in the supply and demand framework. In conventional connotation,
shortage typically refers to an insufficient supply. I adhere to this definition in my discussion
here. A detailed discussion of the confusion caused by terminology can be found in Ingersoll

153
is actually higher in areas such as reading and writing, foreign language, and social science,
where the teacher supply is supposedly abundant. This is the case for Australia, Bulgaria,
Estonia, Hungary, Lithuania, Malaysia, Mexico, Norway, Poland, and the Slovak Republic.
If teacher shortage is the main driver of out-of-field teaching, one would expect to see higher
levels out-of-field teaching in math and science compared to these subject areas, because the
labor demand in math and science is much higher than in other fields in general. One caveat
is that in certain developing countries, because of lack of industrial development, employment
opportunities in manufacturing and the service sector do not generate sufficient demand. In
such cases, maybe teaching is a better job for people who have a college degree, regardless of
their field of training. This scenario is certainly plausible, and a country-by-country analysis
is highly needed to investigate labor demand in particular settings; yet in the above cases,
there is at least a handful of countries which have a highly developed economy (i.e. Australia
and Norway) and I still observed higher levels of out-of-field teaching in the humanities than
math and science.
Furthermore, in the cross-national regression analysis, teacher shortages did not correlate
well with school-level out-of-of-field teaching. This finding holds in the pooled regression
analysis with or without country fixed effects. Certainly there is a role for teacher shortages
in causing out-of-field teaching, but it is not as direct and robust as we have thought before.
More recent empirical evidence from the United States also suggests that we expand our line
of thinking beyond just the teacher preparation pipeline; factors such as a high attrition rate
among math and science teachers causes short-term fluctuation in supply of teachers even if
the pipeline is full (Ingersoll & Perda, 2010).
Second, out-of-field teaching is not a qualification issue. In the past, proposals have been
made to raise the bar of teacher qualification in order to eliminate out-of-field teaching. Some
commentators believe that if every teacher has a college degree, the out-of-field teaching
practice would be effectively curbed. However, empirical evidence does not support this
and Perda (2010).

154
view. In several countries where a bachelor’s degree is almost universal, I still observed
higher levels of out-of-field teaching. More interestingly, out-of-field teaching is even more
prevalent among those who hold a master’s degree in certain countries. This finding suggests
that even if we raise the qualification standard to require teachers to have at least a university
level degree, out-of-field teaching will likely still persist in the classroom. On the other hand,
one could argue that if making out-of-field assignments is inevitable, having someone with
higher levels of education is better than those with lower levels of education. I do not think
we have sufficient information to differentiate between the effectiveness among out-of-field
teachers at this point of time.
These findings lead me to draw the third policy implication: out-of-field teaching is a
school issue. In the cross-country analysis, I showed that giving schools more autonomy to
make decisions, in terms of staffing, deciding the salary schedule, and allocating the budget
for professional development, is correlated with lower levels of out-of-field teaching. The effect
is not trivial. On average, one standard deviation increase in school autonomy will lead to 30
percent less out-of-field teaching from the baseline. Thus, by giving schools more freedom in
making staffing decisions, and deciding on teachers’ salary increases in particular, might be
a feasible way to address out-of-field teaching. Yet because school autonomy is most likely
endogenous within a particular country setting, the effect is highly dependent on national
context. Further analysis is needed to figure out the specific policy lever that can be used
alongside giving schools more autonomy.
Another line of thinking is to look beyond the pure number of out-of-field teachers,
because it is also a distributional issue. The exercise of variance partitioning shows large
between-school variations in the levels of out-of-field teaching across countries. What are
the policy options left to contain it? One thing this analysis confirmed is that out-of-field
teaching is closely related to other forms of teacher sorting mechanisms across schools. In
many countries, out-of-field teachers are more likely to teach in rural schools, small schools,
and low-SES schools. Policymakers should look beyond their national average and closely

155
study disparities between different types of schools. Typically these types of schools’ ability
to attract and retain high quality teachers are much weaker than large city schools, where
not only is the living standard higher, these schools are also close to a large labor market,
hence job opportunities are better. The negative effect of being a small school is perhaps
innate to its nature. To maintain certain economies of scale, teachers are typically required
to teach multiple subjects at the same time. Of course, one simple (and naive) solution is to
limit the number of small schools and mandate a certain minimum size. Yet policymakers
should be cautious of additional costs such as increased student transportation and safety
concerns, because rural and village schools are often located in sparse populated areas as
well.55 In general, the playing field is not equal in terms of teacher staffing. A closer study
of those countries which can ensure a more equal distribution of teachers, such as Bulgaria
and Korea, is essential.
Another implication can be drawn from the close relationship between school SES and
levels of out-of-field teaching. On the one hand, school SES certainly correlates with other
factors such as financial resources and school location. So the correlation between school SES
and out-of-field teaching reflects the previously stated relationship between various forms of
teacher sorting and school types. Without longitudinal data, it is difficult to isolate the
true causal effect of high school SES. On the other hand, qualitative studies have shown
that parental involvement in high-SES schools is much higher (Lareau, 2000). Not only do
highly educated parents choose schools with better teachers, they will also not hesitate to
voice their concern about teachers once their children get in. Parental pressure may play
a role in affecting a school’s staffing decisions. There is emerging international evidence
suggesting that parental involvement and choice are likely to affect schools on a series of
organizational issues, including teacher assignment (Duflo, Dupas, & Kremer, 2011; Pop-
Eleches & Urquiola, 2011). Based on the difference in levels of out-of-field teaching between
55 Also considering these “hard costs” such as school buses, dormitories, dealing with out-
of-field teaching is a financially more acceptable issue and I will not be surprised that it gets
sidelined on the policy agenda.

156
high-SES and low-SES classrooms, we can assume that the notion of out-of-field teachers
may signal a certain degree of teacher quality to parents. Of course there are could be other
confounding factors that I did not observe in the TALIS data but that also played a role
that is associated with within-school teacher sorting. For example, within-school sorting of
students based on their achievement level may well lead to sorting of teachers, with better
qualified techers being assigned to high-achieving classrooms. The dynamics on which these
choices are made is not clear at this time.
From a teacher’s perspective, there is also a role for policy intervention. As I showed
earlier, in several countries younger and inexperienced teachers are more likely to teach out-
of-field. We know that they are at the lower end of the school power hierarchy. They do not
have a lot of say in the teaching assignment process because they are junior, and the data
also suggests that they do not have long contracts to secure their position in schools. This
poses a big challenge from a teacher development perspective, because the first couple of
years of a teacher’s career are vital, both intellectually and professionally (Jensen, Sandoval-
Hernandez, Knoll, & Gonzalez, 2012; Johnson, Kardos, Kauffman, & Liu, 2004). If they
cannot get a good opportunity to teach something for which they have training, how can we
expect them to develop sufficient teaching skills? Actions must be taken to stop assigning
out-of-field positions to these junior faculties.
Another aspect for possible policy innovation is on-the-job development for out-of-field
teachers. Surprisingly, almost no country provides additional support for out-of-field teach-
ers. If we assume that out-of-field teachers lack certain abilities or skills prior to joining the
teaching force, what kind of remedial policies can be in place to help them grow? Literature
suggests that providing mentors and creating a supportive induction program can have a
positive effect on teaching skills and teachers’ retention rate (Rockoff, 2008; Smith & Inger-
soll, 2004). Using various forms of professional development is also a good way to help these
teachers learn. Unfortunately, I did not observe any quantitative difference in professional
development opportunity, appraisal, or collegial collaboration received by out-of-field and

157
in-field teachers. That is to say, they are treated just like any other teachers. Of course,
not to stigmatize out-of-field teachers is in itself a good thing, yet if out-of-field teachers do
not receive extra vital help, it is not reasonable to expect them to perform as well as in-field
teachers. One certainly can argue that if out-of-field teaching is mostly a temporary job,
then all these additional sources of support can be deemed unnecessary. There is certain
value in this argument. If out-of-field teachers are just regular teachers who are assigned to
cover up short-term vacancies, such practice is more justifiable. However the data suggests
that in several countries many of out-of-field teachers are working on part-time or short term
basis. It is unsound if these personnel are just used in a disposable fashion without providing
any professional opportunities (including skills development) to them. Up to this point, I do
not have any data to investigate whether out-of-field teaching indeed does just occur in the
short run. This is an important direction for future research.
With regard to this latter matter, currently there is no productivity measure of teachers
available to me. It may seem subjective to evaluate whether out-of-field teachers are less
effective teachers. Indirect evidence suggests that out-of-field teachers spend less time on
teaching activities, although their approach does not seem to be any different. This is a
future area of research, to use a meaningful measure to compare the effectiveness of out-of-
field teachers.

6.3 Limitation of this study

6.3.1 Limitation in conceptualization of out-of-field teaching

Conceptually, the study of out-of-field teachers offers a narrowly defined policy question:
to what degree does teachers’ subject matter training related to the content he/she is teach-
ing? The fidelity of this research focus is based on the reforms that took place in colleges
or schools of education around world. There are growing contentions and dissatisfaction
toward teacher education since the 1960s (Conant, 1963; Koerner, 1963). The dissatis-

158
faction is based on the fact that teacher education at that time was considered generally
weak in preparing teachers’ subject matter competency. These critique led to a series of
reforms in teacher education programs and teacher certification systems (Sedlak, 2008).
In developed countries such as the United States and Britain, this change occurred in the
second half of the 20th century (Sedlak, 2008; Judge et al., 1994). Yet for a vast majority
of developing nations, this is an ongoing process. With the aid of international develop-
ment agencies, university-based teacher education programs mushroomed in a wide range of
countries (Luschei & Chudgar, 2011).
Today the idea that a properly trained teacher should have at least a college major in the
subject matter has been largely accepted by the public and research community. However, we
should remember that obtaining a major in specific subject matter is only a crude instrument
to measure competency. For instance, in the United States, there are about 1,300 teacher
education programs that offer basic training for future teachers. The quality of these future
teachers, who will all get a major in subject area, varies dramatically. The United States
may be an extreme example of a highly decentralized teacher education system, but it is not
hard to agree that a major or degree in the subject matter offers a minimum competency
measure at best.
The limitation of the out-of-field measure reflects the current challenge to reach a con-
sensus on teacher quality. One promising approach is to use teacher knowledge as a measure
of teacher quality. There are growing efforts that try to measure teacher knowledge based
on Lee Shulman’s (1986; 1988) characterization of the knowledge of teaching. However, at
current stage such study either focuses on a single subject matter or is too small in scale
to be generalized to the national level. There are also challenging measurement issues to be
resolved in developing such a measure. With the development of measurement and statistical
technology, we might soon be able to use that measure to replace the out-of-field measure.
The change in research focus may also help us to focus on the substantive skills of teachers
rather than look at teacher quality at the instrumental level.

159
 Lastly, the theoretical foundation of teacher labor markets has not reached a compre-
hensive stage, especially under international settings. In this dissertation, I explore some
plausible mechanisms that might drive out-of-field teaching. However, as described ear-
lier, national context largely determines the composition and characteristics of teacher labor
force. We are yet to understand what factors can explain out-of-field teacher compositions
across nations. Other unobservable factors such as country-level regulation and adminis-
trative regimes are part of the national context factor. The fixed-effects approach I used
accounted for these unobservables. However I was unable to explain why nations differ from
one another in terms of out-of-field teacher attributes and distribution. To answer the why
question lies beyond the scope of analyzing the TALIS dataset. It requires a deep under-
standing of the contextual knowledge such as the political economy and other social forces.
The lack of country-specific knowledge on teacher labor markets also limits the policy rec-
ommendation I can make. As keen readers will find, several of the explanations offered are
based on extrapolation rather than evidence grounded in solid theoretical foundation. I see
this as a major limitation of this current study. Some of these limitations could be overcome
by detailed country-by-country analysis. I leave this task for the next project.

6.3.2 Limitation of data

At the operational level, there are several issues related to TALIS 2008 data. First
of all, as described in Table 3.2, in at least seven countries, the primary sampling unit
(school) is chosen according to the size of student body per school instead of number of
teachers per school. Ideally in order to ensure that first stage of sampling is truly nationally
representative, schools should be sorted based on number of teachers and a systemic random
sampling should be conducted.56 Because not all the countries collect information on the
number of teachers at each school but usually the number of students are readily available,
56 For detailed procedure of systematic random sampling, please refer to OECD (2010,
p. 59)

160
it then became a practical choice to use number of students as measurement of size (MOS).
Yet the consequences of using the number of students as MOS is not clear. If in one country
the student-teacher ratio is uniform across schools, then it makes no difference which MOS
measure is used to sample the primary sampling unit (school). However, such a scenario
is certainly not realistic in most countries where the student-teacher ratio could vary from
school to school. Then the probability of how schools are sampled might differ depending
which MOS is used, and consequently affect the representativeness of the final teacher sample.
In the current international education literature, there is no empirical work that examines the
implications of using different MOS measures, which casts some doubts on the generalizability
of the findings using international database such as TALIS 2008.
Second, the out-of-field definition used by TALIS 2008 is not directly comparable with
the ones used in previous literature as described in Section 2.2.1. The out-of-field measure
in TALIS 2008 is close to the main assignment measure. However due to the wording of
the survey item, individuals could have different interpretations of the out-of-field question.
For instance, someone with high school-level academic training could justify him/herself
as having sufficient “academic training.” By using this measure, the out-of-field estimates
probably reflects a under-estimated out-of-field teaching level because of a relaxed definition
of academic training. The true levels of out-of-field teaching might be higher, as suggested
by estimates from other international databases such as TIMSS.
Third, this study used cross-sectional data. The benefit of this dataset is to allow re-
searchers to look across a diverse group of countries. The limitation is that there is no
easy way to establish a causal link between the factors I described in previous chapters and
out-of-field assignment. At best the findings in this dissertation is associational instead of
causal. To gauge the casual effect, one would need to rely either on experimental study or
on longitudinal data. These types of studies are more often found in a single-nation setting.
There are many things I would like to have done in this study but could not.
Fourth, since TALIS is not specifically designed for the purpose of finding attributes of

161
out-of-field teachers, the actual sample of out-of-field teachers used in the analysis was small,
averaging 50 teachers per country. This is not a satisfactory sample size by any means if one
thinks about generalizing to all the math and science teachers in the entire country. This
sample size essentially limited the statistical techniques I eventually employed.
Fifth, since this study only uses the mathematics and science teacher sub-samples,
whether or not the findings derived in this study could be applied to other segments of
the teacher labor force and teachers who work in private schools remain to be examined.
Lastly, I also would like to point out that the grade sample varies substantially across
nations depending on how each nation defines “lower secondary education” (Table 3.2). Cur-
rently in the TALIS sample, teachers from various grades are pooled together and there is
no way to differentiate cross-grade variation in out-of-field teaching. Thus it is not clear
whether the out-of-field teaching phenomenon happens across several grades or only applies
to certain levels in schools.

6.4 Recommendation for future TALIS studies

At the time of writing, the next round of TALIS (2013) is well underway. As an impor-
tant component in today’s international education data collection efforts, TALIS makes a
unique contribution to our understanding of teachers’ work. As demonstrated in this disser-
tation, TALIS has much potential to provide scientifically valid evidence for researcher and
policymakers. In this section, I make several recommendations for future waves of TALIS
based on my understanding of TALIS data.
First of all, the current TALIS adopts a design that collects information from all the
teachers within each school. Such a design is beneficial in the sense that it sketches a big
picture of teachers across subject matter areas. Yet the limitation of such an approach is
that it does not allow researchers to look specifically at teachers in a particular subject field,
which might be of substantive interest to researchers. For instance, research on mathematics
teachers has generated much interest. In many countries, math teachers could either have a

162
math major, an education major, or an education minor in his/her teacher preparation. It
would be of substantive interest for researchers to grasp a deep understanding of how math
teachers are prepared across countries. To obtain such information, we will need subject-
specific questions for math teachers. Such survey questions are not implemented in a general
survey protocol that covers all teachers. I recommend that TALIS synchronize its subject
matter coverage with the PISA assessment. For example, in PISA 2012, mathematics will
be the major domain of assessment, and TALIS 2013 could also focus on math teachers to
provide supplemental information. Under this scenario, TALIS should over-sample math
teachers in each school and design separate math teacher survey items. Each future wave
of TALIS could focus on teachers from one or more particular subject fields. Such survey
items could be implemented alongside general questionnaires targeting teachers from other
subject fields. This strategy will have two benefits. It would certainly give researchers who
design the teacher questionnaire more flexibility because they are not limited by sampling
teachers from all the subject matters. Additionally, subject matter focus could substantially
improve the sample size issue I raise in the previous section. With more subject-specific
teachers sampled, the precision of estimates could be greatly improved.
Secondly, TALIS could be a good route to gauge information about teacher education.
It would be ideal to obtain more information on how teachers are trained. Right now the
measure of this kind is crude at best; researchers would benefit from knowing the type of
certificate one received and the type of courses one took in teacher training institutions. For
example, instead of asking teachers about their academic training in general, it would be
beneficial if researchers could gather information about the major and minor each teacher had
previously received and the type of teacher education program they attended. Alternatively,
in recent years assessments of teacher knowledge have emerged in several subject fields. It is
a very effective way to directly evaluate the degree to which teachers master different types
of knowledge at various stages of their career.
Thirdly, TALIS could incorporate more administrative information on teachers. Because

163
this round of TALIS specifically targeted teachers’ professional development opportunities,
many other aspects of teachers’ lives and careers were not included in the data. Such informa-
tion would include the number of assignments in each school, duration of each assignment,
number of classes taught in one semester, grade taught, and previous number of schools
worked. This information could help to understand teachers’ career choices and sustainabil-
ity, which is vital for teacher labor market research. In the future, I would like to see how
long teachers stay in the profession and for what reasons they might leave it. Other features,
such as their feeling of inclusion in the professional community in school, would also be good.
We are yet unaware of the motivations and reactions of out-of-field teachers. To be able to
obtain more detailed information in that regard would be beneficial.
Lastly, TALIS could learn from other single country teacher databases such as the US’s
SASS. As a matured data collection program, SASS has accumulated much experience in
designing survey items and sampling strategies. For instance, SASS has four different types
of out-of-field measures that TALIS is currently lacking. To learn from other successful
data collection programs not only improves the overall data quality, it also helps to link
the potential research outcome with previous research findings by using the same indicators.
Because TALIS is conducted every five years, it will thus accumulate longitudinally compa-
rable statistics to other single-country findings. Of course doing so is not without challenge.
Single-country survey projects are often derived within the particular country context. Some
of the survey items might make sense in one country but not for others. An international
project like TALIS is not an extension of single-country study, but it has to stay relevant
to the participating countries to serve their educational needs. This is no easy task by any
means.
Despite these challenges, as a newcomer in the international education assessments,
TALIS is an ideal source of information to provide researchers with well-rounded evidence
about teachers’ life and work. With growing interests in improving teacher quality and
understanding teacher labor markets, TALIS will continue to grow and evolve. I hope my

164
recommendations will help to improve this new venture.

165
APPENDIX

166
 Appendix A
TECHNICAL NOTES

A.1 Statistical computation

The analytical work of this dissertation is conducted in Stata 12. To account for sampling
weights, the mean statistics are estimated under the svy brr environment. I also used areg
and glm procedure to estimate fixed-effects models.
For multilevel models, I first calculate sampling weights for each teacher within school.57
Then I rescaled the weights using the pwigls command written by Chantala et al. (2011). Al-
though Stata offers increasing flexibility in estimating multilevel mixed models in its xtmixed
command, there are still some restrictions. For binary data with multiple sampling weights, I
used the gllamm procedure developed by Rabe-Hesketh and Skrondal (2008). The estimates
are extracted and formatted using the esttab procedure developed by Jann (2007).
During the process of writing, I benefited much from Long’s (2008) method of organizing
and automating the analysis process. The charts and graphs used in this dissertation are
mostly generated using Stata 12. The methods are described in Mitchell (2008).

A.2 Typesetting

This dissertation is first written in Microsoft Word and then ported to LATEXfor type-
setting. I used the MSU Thesis class58 designed by Prof. Alan Munn from MSU. Another
package of his, csv2Latex, is also used to convert tables from estimation results generated
in Stata by esttab. The texts used the default Computer Modern font family. Most figures
used All the figures used Computer Modern font family with the except of those country
profiles. I used Calibri font for these graphs.

57 This is achieved by dividing the SCHWGT from TCHWGT.

58 See https://www.msu.edu/∼amunn/latex/msu-thesis.zip for detail.

167
 Appendix B
ADDITIONAL TABLES

168
Table B.1: Number of schools by the size of teaching force in math and science in the TALIS sample

Country Number of teachers per school Total schools

1 2 3 4 5 6 7 8 9 10 11 12
Australia 6 7 16 24 17 5 6 0 0 0 0 0 81
Austria 5 29 31 58 41 22 15 3 1 1 0 0 206
Brazil 20 31 35 45 42 41 22 9 3 5 1 0 254
Bulgaria 9 15 17 28 32 34 26 17 9 5 0 0 192
Denmark 6 11 16 11 11 6 10 2 3 0 0 0 76
Estonia 15 30 42 36 30 23 6 4 1 0 0 0 187
Hungary 6 29 26 37 28 21 6 7 0 1 0 0 161
Korea 4 11 6 30 44 13 1 1 0 0 0 0 110
Lithuania 9 26 50 66 37 10 2 0 0 0 0 0 200
Malaysia 3 5 33 53 63 42 8 0 1 0 0 0 208
Malta 0 1 1 16 10 1 1 0 0 0 0 0 30
Mexico 4 14 19 20 27 30 22 7 2 3 0 0 148
Netherlands 2 6 6 8 7 0 0 0 0 0 0 0 29
Norway 15 24 19 14 13 21 10 3 4 5 2 2 132
Poland 2 5 25 44 38 28 13 2 1 0 0 0 158
Portugal 1 4 9 23 40 28 20 4 0 0 0 0 129
Slovak Rep 3 18 23 26 34 16 23 8 4 1 1 0 157
Slovenia 4 21 27 38 30 30 12 5 8 1 0 0 176
Spain 6 6 20 30 29 36 9 1 0 0 0 0 137
Turkey 11 25 19 42 23 16 2 1 0 0 0 0 139
Belgium (F) 9 19 21 19 12 2 0 0 0 0 0 0 82

169
REFERENCES

170
 REFERENCES

Akiba, M., & LeTendre, G. K. (2009). Improving teacher quality: The U.S. teaching force
in global context. New York and London: Teachers College Press.

Akiba, M., LeTendre, G. K., & Scribner, J. P. (2007). Teacher quality, opportunity gap, and
national achievement in 46 countries. Educational Researcher , 36 (7), 369-387. doi:
10.3102/0013189X07308739

Ammermueller, A., & Pischke, J.-S. (2009). Peer effects in European primary schools:
evidence from the Progress in International Reading Literacy Study. Journal of Labor
Economics, 27 (3), 315-338. doi: 10.1086/603650

Andrabi, T., Das, J., & Khwaja, A. I. (2008). A dime a day: The pssibilities and limits
of private schooling in Pakistan. Comparative Education Review , 52 (3), 329–355. doi:
10.1086/588796

Australian Education Union. (2006). AEU National Beginning Teacher Survey Results 2006.
Retrieved from http://www.aeufederal.org.au/Publications/Btsurveygph.pdf

Ballou, D., & Podgursky, M. (1996). Reforming teacher training and recruitment. Govern-
ment Union Review , 17 (4), 1-53.

Barber, M., & Mourshed, M. (2007). How the world’s best-performing school sys-
tems come out on top. London: McKinsey & Company. Retrieved from
http://mckinseyonsociety.com/downloads/reports/Education/Worlds_School
_Systems_Final.pdf

Bobbitt, S. A., & McMillen, M. M. (1994). Qualifications of the public school teacher
workforce: 1988 and 1991. Washington, DC: National Center for Education Statistics.
Retrieved from http://nces.ed.gov/pubsearch/pubsinfo.asp?pubid=95665

Boe, E. E., Shin, S., & Cook, L. H. (2007). Does teacher preparation matter for beginning
teachers in either special or general education? The Journal of Special Education,
41 (3), 158-170. doi: 10.1177/00224669070410030201

Cavanagh, S. (2011). Personnel costs prove tough to contain. Education Week , pp. 26-31. Re-
trieved from http://www.edweek.org/ew/articles/2011/01/13/16personnel.h30
.html?tkn=XRTF5xExaoNFwryKYzmFZgQwRf1f4tDHzQa8&print=1

Chantala, K., Blanchette, D., & Suchindran, C. M. (2011). Software to compute sampling
weights for multilevel analysis (Tech. Rep.). Chapel Hill, NC: Carolina Population Cen-
ter. Retrieved from http://www.cpc.unc.edu/research/tools/data_analysis/ml
_sampling_weights/Compute%20Weights%20for%20Multilevel%20Analysis.pdf

171
Conant, J. B. (1963). The education of American teachers. New York: McGraw-Hill.

Copperman, S., & Klagholz, L. (1985). New Jersey’s alternate route to certification. Phi
Delta Kappan, 66 (10), 691-695.

Credentialing and Certificated Assignments Committee. (2008). Assignment monitoring of

certificated employees in California by county offices of education 2003-2007, A report
to the legislature. Sacramento, CA.

Darling-Hammond, L., Berry, B., & Thoreson, A. (2001). Does teacher certification matter?
Evaluating the evidence. Educational Evaluation and Policy Analysis, 23 (1), 57-77.
doi: 10.3102/01623737023001057

Darling-Hammond, L., Holtzman, D. J., Gatlin, S. J., & Heilig, J. V. (2005). Does teacher
preparation matter? Evidence about teacher certification, Teach for America, and
teacher effectiveness. Education Policy Analysis Archives, 13 (42), 13-48.

Duflo, E., Dupas, P., & Kremer, M. (2011). Peer effects, teacher incentives, and the impact
of tracking: Evidence from a randomized evaluation in Kenya. American Economic
Review , 101 (August), 1739-1774. doi: 10.1257/aer.101.5.1739

Education Week. (2008). Quality counts 2008. Retrieved from http://www.edweek.org/

rc/articles/2008/11/12/sow1112.h27.html?qs=out-of-field

Eurydice. (2002). The teaching profession in Europe (profile, trends and con-
cerns) report 2, supply and demand, general lower secondary education. Brus-
sels: The Information Network on Education in Europe, European Com-
mission. Retrieved from http://eacea.ec.europa.eu/education/eurydice/
documents/thematic_reports/archives/034EN.zip

Goldhaber, D. D., & Brewer, D. J. (2000). Does teacher certification matter? High school
teacher certification status and student achievement. Educational Evaluation and Pol-
icy Analysis, 22 (2), 129-145. doi: 10.3102/01623737022002129

Gorard, S., See, B. H., Smith, E., & White, P. (2007). What can we do to strengthen the
teacher workforce? International Journal of Lifelong Education, 26 (4), 419-437. doi:
10.1080/02601370701417194

Guarino, C. M., Santibanez, L., & Daley, G. A. (2006). Teacher recruitment and retention:
A review of the recent empirical literature. Review of Educational Research, 76 (2),
173-208. doi: 10.3102/00346543076002173

Hanushek, E. A., & Woessmann, L. (2010). The economics of international differences in

educational achievement. In E. A. Hanushek, S. Machin, & L. Woessmann (Eds.),
Handbook of the economics of education volume 3 (p. 89-200). Amsterdam: North
Holland: Elsevier.

172
Hawk, P. P., Coble, C. R., & Swanson, M. (1985). Certification: It does matter. Journal of
Teacher Education, 36 , 13-15. doi: 10.1177/002248718503600303

Hechinger, F. M. (1985, October 8). "Dirty, little secret" of unlicensed teachers. The
New York Times. Retrieved from http://www.nytimes.com/1985/10/08/science/
about-education-dirty-little-secret-of-unlicensed-teachers.html?scp=
23&sq=Albert+Shanker&st=nyt

Hill, H. C., Ball, D. L., Blunk, M., Geoffney, I. M., & Rowan, B. (2007). Validating the
ecological assumption: The relationship of measure scores to classroom teaching and
student learning. Measurement: Interdisciplinary Research & Perspective, 5 (2), 107-
118. doi: 10.1080/15366360701487138

Hill, H. C., Rowan, B., & Ball, D. L. (2005). Effects of teachers’ mathematical knowledge
for teaching on student achievement. American Educational Research Journal , 42 (2),
371-406. doi: 10.3102/00028312042002371

Holland, P. W. (1986). Statistics and causal inference. Journal of the American Statistical
Association, 81 (396), 945-960. doi: 10.2307/2289069

Imig, D. G., & Imig, S. R. (2008). From traditional certification to competitive certifi-
cation: A twenty-five year retrospective. In M. Cochran-Smith, S. Feiman-Nemser,
D. J. McIntyre, & K. E. Demers (Eds.), Handbook of research on teacher education:
Enduring questions in changing contexts (3rd ed., p. 886-907). New York: Routledge.

Ingersoll, R. M. (1999). The problem of underqualified teachers in American secondary

schools. Educational Researcher , 28 (2), 26-37. doi: 10.3102/0013189X028002026

Ingersoll, R. M. (2002). Out-of-Field teaching, educational inequality, and other organization

of schools: An exploratory analysis. Seattle, WA: Center for the Study of Teaching
and Policy.

Ingersoll, R. M. (2003). Out-of-Field teaching and the limits of teacher policy. Philadelphia,
PA: Center for the Study of Teaching and Policy.

Ingersoll, R. M. (2004). Why some schools have more underqualified teachers than others?
In D. Ravitch (Ed.), Brookings papers on education policy (p. 45-88). Washington,
DC: Brookings Institution Press. Retrieved from http://repository.upenn.edu/
gse_pubs/144/

Ingersoll, R. M. (2005). The problem of underqualified teachers: A sociological perspective.

Sociology of Education, 78 (2), 175–178. doi: 10.1177/003804070507800206

Ingersoll, R. M. (2006). Who controls teachers’ work? Power and accountability in America’s
schools. Cambridge, MA: Harvard University Press.

Ingersoll, R. M., & Perda, D. (2010). Is the supply of mathematics and science teachers

173
 sufficient? American Educational Research Journal , 47 (3), 563-594. doi: 10.3102/
0002831210370711

Jann, B. (2007). Making regression tables simplified. The Stata Journal , 7 (2), 227–244.
Retrieved from http://www.stata-journal.com/article.html?article=st0085

Jensen, B., Sandoval-Hernandez, A., Knoll, S., & Gonzalez, E. J. (2012). The experi-
ence of new teachers: Results from TALIS 2008. OECD Publishing. doi: 10.1787/
9789264120952-en

Jerald, C. D., & Boser, U. (2000). Setting policies for new teachers. Retrieved from http://
www.edweek.org/media/ew/qc/archives/QC00full.pdf

Jerald, C. D., & Ingersoll, R. (2002). All talk, No action: Putting an end to out-of-field
teaching. Washington, DC: Education Trust.

Johnson, S. M., Kardos, S. M., Kauffman, D., & Liu, E. (2004). The support gap: New
teachers’ early experiences in high-income and low-income schools. Education Policy
Analysis Archives, 12 (61), 1-23.

Judge, H., Lemosse, M., Paine, L., & Sedlak, M. W. (1994). The university and the teachers:
France, the United States, England. Wallingford, Oxfordshire: Triangle Books.

Koerner, J. D. (1963). The miseducation of American teachers. Boston: Houghton Mifflin.

Ladd, H. F. (2007). Teacher labor markets in developed countries. The Future of Children,
17 (1), 201-217.

Lankford, H., Loeb, S., & Wykoff, J. (2002). Teacher sorting and the plight of urban schools:
A descriptive analysis. Educational Evaluation and Policy Analysis, 24 (1), 37–62. doi:
10.3102/01623737024001037

Lareau, A. (2000). Home advantage: Social class and parental intervention in elementary
education (2nd ed.). Rowman & Littlefield Publishers.

Lazear, E. P. (2003). Teacher incentives. Swedish Economic Policy Review , 10 , 179-214.

Lee, C., & Krapfl, L. (2002). Teaching as you would have them teach : An effective ele-
mentary science teacher preparation program. Journal of Science Teacher Education,
13 (3), 247-265. doi: 10.1023/A:1016521201095

Lewis, L., Parsad, B., Carey, N., Bartfai, N., Farris, E., & Smerdon, B. (1999). Teacher
quality: A report on the preparation and qualifications of public school teachers.
Washington, DC: National Center for Education Statistics. Retrieved from http://
nces.ed.gov/pubs99/1999080.pdf

Long, J. S. (2008). The Workflow of data analysis using Stata. College Station, TX: Stata

174
 Press.

Lowell, B. L., & Salzman, H. (2007). Into the eye of the storm: Assessing the evidence on
science and engineering education, quality, and workforce demand. Washington, DC:
the Urban Institute.

Luschei, T. F., & Chudgar, A. (2011). Teachers, student achievement and national income:
A cross-national examination of relationships and interactions. Prospects, 41 (1). doi:
10.1007/s11125-011-9213-7

Lynn, F. B., & Keesler, V. (2008). Beyond compliance: Teacher licensure and endorsement
in the state of Michigan, technical report No. 3.1. East Lansing, MI: Michigan State
University.

Lynn, F. B., Keesler, V., Diember, M., Guarino, C. M., Jones, N., Wang, Q., . . . Zhou,
Y. (2007). Beyond compliance: Descriptive characteristics of public school teachers in
Michigan. East Lansing, MI: Michigan State University.

Martin, M. O., Mullis, I. V., Foy, P., Olson, J. F., Erberber, E., Preuschoff, C., & Galia, J.
(2008). TIMSS 2007 international science report: Findings from IEA’s Trends in In-
ternational Mathematics and Science Study at the fourth and eighth Grades. Chestnut
Hill, MA: TIMSS & PIRLS International Study Center, Boston College.

Martin, M. O., Mullis, I. V., Foy, P., Olson, J. F., Preuschoff, C., Erberber, E., . . . Galia, J.
(2008). TIMSS 2007 international mathematics report: Findings from IEA’s Trends in
International Mathematics and Science Study at the fourth and eighth grades. Chestnut
Hill, MA: TIMSS & PIRLS International Study Center, Boston College.

Martin, M. O., Mullis, I. V. S., Gonzalez, E. J., & Chrostowski, S. J. (2004). TIMSS 2003
international science report. Chestnut Hill, MA: TIMSS & PIRLS International Study
Center, Boston College.

McEwan, P. J., & Carnoy, M. (2000). The effectiveness and efficiency of private schools in
Chile’s voucher system. Educational Evaluation and Policy Analysis, 22 (3), 213-239.
doi: 10.3102/01623737022003213

Mello, V. B. D., & Broughman, S. P. (1996). SASS by state 1993-94 Schools and Staffing
Survey : Selected results. Washington, DC: National Center for Education Statistics.
Retrieved from http://nces.ed.gov/pubsearch/pubsinfo.asp?pubid=96312

Mitchell, M. N. (2008). A visual guide to Stata graphics (2nd ed.). College Station, Texas:
Stata Press.

Morton, B. A., Peltola, P., Hurwitz, M. D., Orlofsky, G. F., Strizek, G. A., & Gruber,
K. J. (2008). Education and certification qualifications of departmentalized public high
school-level teachers of core subjects: Evidence from the 2003-04 Schools and Staffing

175
 Survey. Washington, DC: National Center for Education Statistics. Retrieved from
http://nces.ed.gov/pubsearch/pubsinfo.asp?pubid=2008338

Mullis, I. V. S., Martin, M. O., Gonzalez, E. J., & Chrostowski, S. J. (2004). TIMSS 2003
international mathematics report. Chestnut Hill, MA: TIMSS & PIRLS International
Study Center, Boston College.

Muralidharan, K., & Kremer, M. (2008). Public and private schools in rural india. In School
choice international: Exploring public-private partnerships (p. 91-110). Princeton, NJ:
The MIT Press.

National Academy of Sciences, National Academy of Engineering, & Institute of Medicine.

(2007). Rising above the gathering storm: Energizing and employing America for a
brighter economic future. Washington, DC: The National Academies Press.

National Commission on Teaching & America’s Future. (1996). What matters most: Teach-
ing for America’s future. Washington, DC. Retrieved from http://nctaf.org/
wp-content/uploads/2012/01/WhatMattersMost.pdf

National Comprehensive Center for Teacher Quality. (2009). Key issue: Recruiting sci-
ence, technology , engineering, and mathematics (STEM) teachers. Washington, DC:
National Comprehensive Center for Teacher Quality.

National Research Council. (2002). Learning and understanding: Improving advanced study
of mathematics and science in U.S. schools. Washington, DC: The National Academies
Press.

OECD. (2005). Teachers matter: Attracting, developing and retaining effective teachers.
Paris: OECD Publishing. doi: 10.1787/9789264018044-en

OECD. (2009). Creating effective teaching and learning environments: First results from
TALIS. Paris: OECD Publishing. doi: 10.1787/9789264068780-en

OECD. (2010). TALIS 2008 technical report. Paris: OECD Publishing. doi: 10.1787/
9789264079861-en

Ohinata, A., & van Ours, J. C. (2011). How immigrant children affect the academic
achievement of native Dutch children (Discussion Paper No. 8718). London: Cen-
tre for Economic Policy Research. Retrieved from http://www.cepr.org/pubs/dps/
DP8718.asp

Papke, L. E., & Wooldridge, J. M. (1996). Econometric methods for fractional response
variables with an application to 401(k) plan participation rates. Journal of Applied
Econometrics, 11 (6), 619-632. doi: 10.1002/(SICI)1099-1255(199611)11:6<619::AID
-JAE418>3.0.CO;2-1

Pfeffermann, D., Skinner, C. J., Holmes, D. J., Goldstein, H., & Rasbash, J. (1998). Weight-

176
 ing for unequal selection probabilities in multilevel models. Journal of the Royal Sta-
tistical Society. Series B (Statistical Methodology), 60 (1), 23-40. doi: DOI:10.1111/
1467-9868.00106

Pop-Eleches, C., & Urquiola, M. (2011, March). Going to a better school: Effects and behav-
ioral responses (Working Paper No. 16886). National Bureau of Economic Research.
Retrieved from http://www.nber.org/papers/w16886

Psacharopoulos, G. (1987). Public versus private schools in developing countries: Evidence

from Colombia and Tanzania. International Journal of Educational Development, 7 (1),
59-67. doi: 10.1016/0738-0593(87)90007-1

Rabe-Hesketh, S., & Skrondal, A. (2008). Multilevel and longitudinal modeling using Stata
(2nd ed.). College Station, TX: Stata Press.

Raudenbush, S. W., & Bryk, A. S. (2001). Hierarchical linear models: Applications and
data analysis methods (2nd ed.; J. D. Leeuw, Ed.). Thousands Oaks, California: Sage
Publications, Inc.

Rockoff, J. E. (2008, March). Does mentoring reduce turnover and improve skills of new
employees? evidence from teachers in new york city (Working Paper No. 13868). Na-
tional Bureau of Economic Research. Retrieved from http://www.nber.org/papers/
w13868

Rowan, B., Chiang, F.-S., & Miller, R. J. (1997). Using research on employees’ performance
to study the effects of teachers on students’ achievement. Sociology of Education,
70 (4), 256-284.

Rutkowski, L., Gonzalez, E., Joncas, M., & von Davier, M. (2010). International large-scale
assessment data: Issues in secondary analysis and reporting. Educational Researcher ,
39 (2), 142-151. doi: 10.3102/0013189X10363170

Schleicher, A. (2011). Building a high-quality teaching profession: Lessons from around the
world. OECD Publishing. doi: 10.1787/9789264113046-en

Schmidt, W. H., Houang, R. T., Cogan, L., Blömeke, S., Tatto, M. T., Hsieh, F. J., . . .
Paine, L. (2008). Opportunity to learn in the preparation of mathematics teachers: Its
structure and how it varies across six countries. Zdm, 40 (5), 735-747. doi: 10.1007/
s11858-008-0115-y

Schmidt, W. H., Tatto, M. T., Bankov, K., Blomeke, S., Cedillo, T., Cogan, L., . . . Schwille,
J. (2007). The preparation gap: Teacher education for middle school mathematics
in six countries. East Lansing: Michigan State University Center for Research in
Mathematics and Science Education.

Schools and Staffing Survey. (2010). Public school teacher data file, 2007-08. National

177
 Center for Education Statistics. Retrieved from http://nces.ed.gov/surveys/sass/
tables/sass0708_006_t1n.asp

Seastrom, M. M., Gruber, K. J., Henke, R. R., McGrath, D. J., & Cohen, B. A. (2002).
Qualifications of the public school teacher workforce: Prevalence of out-of-field teaching.
Washington, DC: National Center for Education Statistics. Retrieved from http://
nces.ed.gov/pubs2002/2002603.pdf

Sedlak, M. W. (1989). "Let us go and buy a schoolmaster": Historical perspectives on

the hiring of teachers in the United States, 1750-1980. In D. Warren (Ed.), American
teachers: Histories of a profession at work (p. 257-290). New York: Macmillan.

Sedlak, M. W. (2008). Competing visions of purpose, practice, and policy: The history of
teacher certification in the United States. In M. Cochran-Smith, S. Feiman-Nemser,
D. J. McIntyre, & K. E. Demers (Eds.), Handbook of research on teacher education:
Enduring questions in changing contexts (3rd ed., p. 855-885). New York: Routledge.

Shulman, L. S. (1986). Those who understand: Knowledge growth in teaching. Educational

Researcher , 15 (4), 4-14.

Shulman, L. S. (1988). Knowledge growth in teaching: A final report to the Spencer Foun-
dation. Stanford, CA: Stanford University.

Smith, T. M., & Ingersoll, R. M. (2004). What are the effects of induction and mentoring on
beginning teacher turnover? American Educational Research Journal , 41 (3), 681-714.
doi: 10.3102/00028312041003681

UNESCO Institute for Statistics. (2006). Teachers and educational quality: Monitoring
global needs for 2015. Montreal, Canada: UNESCO Institute for Statistics. Retrieved
from http://unesdoc.unesco.org/images/0014/001457/145754e.pdf

U.S. Department of Education. (2002). Meeting the highly qualified teachers challenge: The
secretary’s annual report on Teacher Quality. Washington, DC.

Wilson, S. M., Ball, D. L., Bryk, A. S., Figlio, D. N., Grossman, P., Irvine, J. J., . . . Porter,
A. (2009). Teacher quality: Education policy White paper. Washington, DC: National
Academy of Education.

Youngs, P., & Grogan, E. (2010). United States country report on teacher education. East
Lansing: Michigan State University.

178